atom.xml

<?xml version="1.0" encoding="utf-8"?>
<feed xmlns="http://www.w3.org/2005/Atom">
  <title>李飞阳</title>
  <subtitle>PM、Coder、Data mining</subtitle>
  <link href="/atom.xml" rel="self"/>
  
  <link href="http://feiyang.li/"/>
  <updated>2017-06-08T04:10:47.347Z</updated>
  <id>http://feiyang.li/</id>
  
  <author>
    <name>Wing Lee</name>
    
  </author>
  
  <generator uri="http://hexo.io/">Hexo</generator>
  
  <entry>
    <title>算法实验(二)</title>
    <link href="http://feiyang.li/2017/06/08/algorithm2/index.html"/>
    <id>http://feiyang.li/2017/06/08/algorithm2/index.html</id>
    <published>2017-06-08T04:09:41.000Z</published>
    <updated>2017-06-08T04:10:47.347Z</updated>
    
    <content type="html"><![CDATA[<script src="/assets/js/DPlayer.min.js"> </script><script src="/assets/js/APlayer.min.js"> </script><blockquote>
<p>任务调度问题：在单处理器上具有期限和惩罚的单位时间任务调度问题；平衡树问题：实现3种树中的两种：红黑树，AVL树，Treap树。</p>
</blockquote>
<hr>
<a id="more"></a>
<h1 id="任务调度问题"><a href="#任务调度问题" class="headerlink" title="任务调度问题"></a>任务调度问题</h1><p>任务调度问题就是给定一个有穷单位时间任务的集合S，集合S中的每个任务都有一个截止期限di和超时惩罚wi，需要找出集合S的一个调度，使得因任务误期所导致的总惩罚最小，这个调度也称为S的一个最优调度。</p>
<h2 id="问题描述："><a href="#问题描述：" class="headerlink" title="问题描述："></a>问题描述：</h2><p>在单处理器上具有期限和惩罚的单位时间任务调度问题（课本P239）<br>实验要求：<br>（1）实现这个问题的贪心算法<br>（2）将每个 wi 替换为max{m1,m2…mn}—wi，运行算法比较结果。 </p>
<h2 id="解题思路："><a href="#解题思路：" class="headerlink" title="解题思路："></a>解题思路：</h2><p>1.先将任务按照时间惩罚递减顺序进行排序，<br>2.然后用贪心的思想，尽量把惩罚重的任务先放入待完成队列中。 </p>
<h2 id="代码"><a href="#代码" class="headerlink" title="代码"></a>代码</h2><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div><div class="line">10</div><div class="line">11</div><div class="line">12</div><div class="line">13</div><div class="line">14</div><div class="line">15</div><div class="line">16</div><div class="line">17</div><div class="line">18</div><div class="line">19</div><div class="line">20</div><div class="line">21</div><div class="line">22</div><div class="line">23</div><div class="line">24</div><div class="line">25</div><div class="line">26</div><div class="line">27</div><div class="line">28</div><div class="line">29</div><div class="line">30</div><div class="line">31</div><div class="line">32</div><div class="line">33</div><div class="line">34</div><div class="line">35</div><div class="line">36</div><div class="line">37</div><div class="line">38</div><div class="line">39</div><div class="line">40</div><div class="line">41</div><div class="line">42</div><div class="line">43</div><div class="line">44</div><div class="line">45</div><div class="line">46</div><div class="line">47</div><div class="line">48</div><div class="line">49</div><div class="line">50</div><div class="line">51</div><div class="line">52</div><div class="line">53</div><div class="line">54</div><div class="line">55</div><div class="line">56</div><div class="line">57</div><div class="line">58</div><div class="line">59</div><div class="line">60</div><div class="line">61</div><div class="line">62</div><div class="line">63</div><div class="line">64</div><div class="line">65</div><div class="line">66</div><div class="line">67</div><div class="line">68</div><div class="line">69</div><div class="line">70</div><div class="line">71</div><div class="line">72</div><div class="line">73</div><div class="line">74</div><div class="line">75</div><div class="line">76</div><div class="line">77</div><div class="line">78</div><div class="line">79</div><div class="line">80</div><div class="line">81</div><div class="line">82</div><div class="line">83</div><div class="line">84</div><div class="line">85</div><div class="line">86</div><div class="line">87</div><div class="line">88</div><div class="line">89</div><div class="line">90</div><div class="line">91</div><div class="line">92</div><div class="line">93</div><div class="line">94</div><div class="line">95</div><div class="line">96</div><div class="line">97</div><div class="line">98</div><div class="line">99</div><div class="line">100</div><div class="line">101</div><div class="line">102</div></pre></td><td class="code"><pre><div class="line">#include&lt;iostream&gt;</div><div class="line">#include &lt;algorithm&gt;</div><div class="line">#include&lt;ctime&gt;</div><div class="line">#define n 6</div><div class="line">using namespace std;</div><div class="line">typedef struct &#123;</div><div class="line">    int id;//第几个任务</div><div class="line">    int d;//任务的期限</div><div class="line">    int w;//罚</div><div class="line">&#125;tesk;</div><div class="line"></div><div class="line">bool cmpW(tesk t1, tesk t2) &#123;</div><div class="line">    return t1.w&gt;t2.w;</div><div class="line">&#125;</div><div class="line">bool cmpD(tesk t1, tesk t2) &#123;</div><div class="line">    return t1.d&lt;t2.d;</div><div class="line">&#125;</div><div class="line">void InitNum(tesk t[]) &#123;</div><div class="line">    srand((unsigned)time(NULL));</div><div class="line">    for (int i = 0; i &lt; n; i++) &#123;</div><div class="line">        t[i].id = i + 1;</div><div class="line">        t[i].d = rand() % n + 1;</div><div class="line">        t[i].w = rand() % 30;</div><div class="line">    &#125;</div><div class="line">&#125;</div><div class="line"></div><div class="line">void copy(tesk &amp;t1, tesk &amp;t2) &#123;</div><div class="line">    t1.id = t2.id;</div><div class="line">    t1.d = t2.d;</div><div class="line">    t1.w = t2.w;</div><div class="line">&#125;</div><div class="line"></div><div class="line">int greedy(tesk a[], tesk ta[], int &amp;ans) &#123;</div><div class="line">    int num = 0, i, j;//当前已经完成的任务数量</div><div class="line">    sort(ta, ta + n, cmpW);</div><div class="line">    int fla[9999];</div><div class="line">    memset(fla, 0, sizeof(fla));</div><div class="line">    for (i = 0; i&lt;n; i++) &#123;</div><div class="line">        for (j = ta[i].d; j&gt;0; j--) &#123;</div><div class="line">            if (fla[j] == 0) &#123;</div><div class="line">                fla[j] = 1;</div><div class="line">                break;</div><div class="line">            &#125;</div><div class="line">        &#125;</div><div class="line">        if (j&gt;0) &#123;</div><div class="line">            copy(a[num++], ta[i]);</div><div class="line">            ans -= ta[i].w;</div><div class="line">        &#125;</div><div class="line">    &#125;</div><div class="line">    return num;</div><div class="line">&#125;</div><div class="line"></div><div class="line"></div><div class="line">int main() &#123;</div><div class="line">    tesk ta[n], a[n], tb[n], b[n];</div><div class="line">    printf(&quot;%s&quot;, &quot;正在生成随机数据\n&quot;);</div><div class="line">    InitNum(ta);</div><div class="line">    for (int i = 0; i &lt; n; i++) &#123;</div><div class="line">        copy(tb[i], ta[i]);</div><div class="line">    &#125;</div><div class="line">    printf(&quot;%s&quot;, &quot;生成的数据为\n&quot;);</div><div class="line">    int max_w = -1, sum_w = 0;</div><div class="line">    for (int i = 0; i &lt; n; i++) &#123;</div><div class="line">        printf(&quot;ID: %d，任务的期限: %d，任务的惩罚：%d\n&quot;, ta[i].id, ta[i].d, ta[i].w);</div><div class="line">        sum_w += ta[i].w;</div><div class="line">        if (max_w &lt; ta[i].w) &#123;</div><div class="line">            max_w = ta[i].w;</div><div class="line">        &#125;</div><div class="line">    &#125;</div><div class="line">    int k = greedy(a, ta, sum_w);</div><div class="line">    printf(&quot;任务惩罚为：%d\n&quot;, sum_w);</div><div class="line">    sort(a, a + k, cmpD);</div><div class="line">    printf(&quot;将完成执行的任务按照时间递增顺序排列输出：\n&quot;);</div><div class="line">    for (int i = 0; i&lt;k; i++) &#123;</div><div class="line">        printf(&quot;ID：%d，期限：%d，惩罚：%d\n&quot;, a[i].id, a[i].d, a[i].w);</div><div class="line">    &#125;</div><div class="line">    sort(a, a + k, cmpW);</div><div class="line">    printf(&quot;将完成的任务按惩罚排序；\n&quot;);</div><div class="line">    for (int i = 0; i&lt;k; i++) &#123;</div><div class="line">        printf(&quot;ID：%d，期限：%d，惩罚：%d\n&quot;, a[i].id, a[i].d, a[i].w);</div><div class="line">    &#125;</div><div class="line"></div><div class="line">    printf(&quot;若将每个wi替换为max&#123;m1,m2...mn&#125;—wi，新的数据为\n&quot;);</div><div class="line">    for (int i = 0; i&lt;n; i++) &#123;</div><div class="line">        tb[i].w = max_w - tb[i].w;</div><div class="line">        printf(&quot;ID：%d，期限：%d，惩罚：%d\n&quot;, tb[i].id, tb[i].d, tb[i].w);</div><div class="line">        sum_w += tb[i].w;</div><div class="line">    &#125;</div><div class="line">    k = greedy(b, tb, sum_w);</div><div class="line">    printf(&quot;任务惩罚为：%d\n&quot;, sum_w);</div><div class="line">    sort(b, b + k, cmpD);</div><div class="line">    printf(&quot;将完成执行的任务按照时间递增顺序排列输出：\n&quot;);</div><div class="line">    for (int i = 0; i&lt;k; i++) &#123;</div><div class="line">        printf(&quot;ID：%d，期限：%d，惩罚：%d\n&quot;, b[i].id, b[i].d, b[i].w);</div><div class="line">    &#125;</div><div class="line">    sort(b, b + k, cmpW);</div><div class="line">    printf(&quot;将完成的任务按惩罚排序；\n&quot;);</div><div class="line">    for (int i = 0; i&lt;k; i++) &#123;</div><div class="line">        printf(&quot;ID：%d，期限：%d，惩罚：%d\n&quot;, b[i].id, b[i].d, b[i].w);</div><div class="line">    &#125;</div><div class="line"></div><div class="line">&#125;</div></pre></td></tr></table></figure>
<h2 id="实验截图"><a href="#实验截图" class="headerlink" title="实验截图"></a>实验截图</h2><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170605/190729182.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<h1 id="实现3种树"><a href="#实现3种树" class="headerlink" title="实现3种树"></a>实现3种树</h1><p>实现3种树中的两种：红黑树，AVL树(课本P177)，Treap树</p>
<h2 id="红黑树"><a href="#红黑树" class="headerlink" title="红黑树"></a>红黑树</h2><h3 id="性质"><a href="#性质" class="headerlink" title="性质"></a>性质</h3><ul>
<li>根节点是黑色的</li>
<li>每个节点或红或黑</li>
<li>每个叶节点都是黑色的</li>
<li>红节点的两个子节点都是黑色</li>
</ul>
]]></content>
    
    <summary type="html">
    
      &lt;blockquote&gt;
&lt;p&gt;任务调度问题：在单处理器上具有期限和惩罚的单位时间任务调度问题；平衡树问题：实现3种树中的两种：红黑树，AVL树，Treap树。&lt;/p&gt;
&lt;/blockquote&gt;
&lt;hr&gt;
    
    </summary>
    
    
      <category term="Algorithms" scheme="http://feiyang.li/tags/Algorithms/"/>
    
      <category term="cpp" scheme="http://feiyang.li/tags/cpp/"/>
    
  </entry>
  
  <entry>
    <title>Python mysql表数据和json格式的相互转换</title>
    <link href="http://feiyang.li/2017/05/31/python-mysql-json/index.html"/>
    <id>http://feiyang.li/2017/05/31/python-mysql-json/index.html</id>
    <published>2017-05-31T13:01:18.000Z</published>
    <updated>2017-05-31T13:02:02.907Z</updated>
    
    <content type="html"><![CDATA[<script src="/assets/js/DPlayer.min.js"> </script><script src="/assets/js/APlayer.min.js"> </script><blockquote>
<p>今天碰到了一个比较奇葩的需求，将mysql表数据转换成json格式。我顺便研究了一下，写了个python。</p>
</blockquote>
<hr>
<a id="more"></a>
<ul>
<li>数据库相关的<a href="http://feiyang.li/2017/05/16/databases/index.html">基本操作</a> </li>
<li>如果使用<a href="http://feiyang.li/2017/02/10/python_camp3/index.html">python操作mysql</a></li>
</ul>
<p>要我说，数据库操作学会了之后，会节省很多很多时间。</p>
<h1 id="SQL数据—-gt-json数据"><a href="#SQL数据—-gt-json数据" class="headerlink" title="SQL数据—&gt;json数据"></a>SQL数据—&gt;json数据</h1><figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div><div class="line">10</div><div class="line">11</div><div class="line">12</div><div class="line">13</div><div class="line">14</div><div class="line">15</div><div class="line">16</div><div class="line">17</div><div class="line">18</div><div class="line">19</div><div class="line">20</div><div class="line">21</div><div class="line">22</div><div class="line">23</div><div class="line">24</div><div class="line">25</div><div class="line">26</div><div class="line">27</div><div class="line">28</div><div class="line">29</div><div class="line">30</div><div class="line">31</div><div class="line">32</div><div class="line">33</div><div class="line">34</div><div class="line">35</div><div class="line">36</div><div class="line">37</div><div class="line">38</div><div class="line">39</div><div class="line">40</div><div class="line">41</div><div class="line">42</div><div class="line">43</div><div class="line">44</div></pre></td><td class="code"><pre><div class="line"><span class="comment">#!/usr/bin/env python</span></div><div class="line"><span class="comment">#-*- coding: utf8 -*-</span></div><div class="line"></div><div class="line"><span class="keyword">import</span> MySQLdb</div><div class="line"><span class="keyword">import</span> json</div><div class="line"></div><div class="line"><span class="comment"># 定义查询SQL语句</span></div><div class="line">sql = <span class="string">"SELECT id,NAME,LOCAL,mobile,CreateTime FROM db1.s1;"</span></div><div class="line"></div><div class="line"><span class="comment"># 设置列的别名或者直接用表字段名:（下面的sql含有别名）</span></div><div class="line"></div><div class="line"><span class="comment"># sql="SELECT s.id AS 1_id,s.name AS 2_na,s.local AS 3_lo,s.mobile AS 4_mo,s.CreateTime AS 5_ct FROM db1.s1 s;"</span></div><div class="line"></div><div class="line"></div><div class="line"><span class="comment"># 定义连接MySQL的登录信息（此处以字典形式）</span></div><div class="line">Loginfo = &#123;<span class="string">'USER'</span>:<span class="string">'admin'</span>, <span class="string">'PSWD'</span>:<span class="string">'admin'</span>, <span class="string">'HOST'</span>:<span class="string">'10.10.60.108'</span>, <span class="string">'PORT'</span>:<span class="number">4001</span>&#125;</div><div class="line"></div><div class="line"></div><div class="line"><span class="comment"># Python 连接MySQL</span></div><div class="line">conn=MySQLdb.connect(host=Loginfo[<span class="string">'HOST'</span>],user=Loginfo[<span class="string">'USER'</span>],passwd=Loginfo[<span class="string">'PSWD'</span>],port=Loginfo[<span class="string">'PORT'</span>],charset=<span class="string">'utf8'</span>)</div><div class="line">cur=conn.cursor()</div><div class="line">cur.execute(sql)                        <span class="comment"># 执行SQL查询</span></div><div class="line">data = cur.fetchall()                   <span class="comment"># 查询结果给data。如果执行：print data 显示结果：（（第一行内容），（第二行内容），（第三行内容），（第四行内容））</span></div><div class="line">fields = cur.description               <span class="comment"># 获取查询结果中列的字段名，如果查询SQL中使用别名，此处显示别名。       </span></div><div class="line">cur.close()</div><div class="line">conn.close()</div><div class="line"></div><div class="line"><span class="comment"># Main </span></div><div class="line">column_list = []                        <span class="comment"># 定义字段名的列表</span></div><div class="line"><span class="keyword">for</span> i <span class="keyword">in</span> fields:</div><div class="line">    column_list.append(i[<span class="number">0</span>])　　  <span class="comment"># 提取字段名，追加到列表中</span></div><div class="line"><span class="comment">#print column_list　　　　　　　   # 列表显示结果：['id', 'NAME', 'LOCAL', 'mobile', 'CreateTime']</span></div><div class="line"></div><div class="line"><span class="keyword">with</span> open(<span class="string">'/data/scripts/logdb/json.txt'</span>,<span class="string">'w+'</span>) <span class="keyword">as</span> f:　　<span class="comment"># 打开输出结果文件</span></div><div class="line">    <span class="keyword">for</span> row <span class="keyword">in</span> data:　　　　　　　　　　　　　　　　　　 <span class="comment"># 一次循环，row代表一行，row以元组的形式显示。</span></div><div class="line">        result = &#123;&#125;　　　　　　　　　　　　　　　　　　　<span class="comment"># 定义Python 字典</span></div><div class="line">        result[column_list[<span class="number">0</span>]] = row[<span class="number">0</span>]　　　  　　　　　<span class="comment"># 将row中的每个元素，追加到字典中。　</span></div><div class="line">        result[column_list[<span class="number">1</span>]] = str(row[<span class="number">1</span>])　　　　　　 <span class="comment"># Python字段格式 和json字段格式转换</span></div><div class="line">        result[column_list[<span class="number">2</span>]] = str(row[<span class="number">2</span>]) </div><div class="line">        result[column_list[<span class="number">3</span>]] = row[<span class="number">3</span>]</div><div class="line">        result[column_list[<span class="number">4</span>]] = str(row[<span class="number">4</span>])</div><div class="line">        jsondata=json.dumps(result,ensure_ascii=<span class="keyword">False</span>) <span class="comment"># Python的dict --转换成----&gt; json的object</span></div><div class="line">        f.write(jsondata + <span class="string">'\n'</span>)　　　　　　　　　　　　　<span class="comment"># 写入文件</span></div><div class="line">f.close()</div></pre></td></tr></table></figure>
<p>注意：最后输出结果，列的顺序是随机的。</p>
<p>　　如果想要顺序输出：1.字段起别名，比如：1_xx,2_xx,3_xx…..然后在在json.dumps中利用：sort_keys=True排序，结果就顺序输出。</p>
<p>例如：</p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">print json.dumps(&#123;&quot;c&quot;: 0, &quot;b&quot;: 0, &quot;a&quot;: 0&#125;, sort_keys=True)</div></pre></td></tr></table></figure>
<p>{“a”: 0, “b”: 0, “c”: 0}</p>
<h1 id="json格式—-gt-sql语句"><a href="#json格式—-gt-sql语句" class="headerlink" title="json格式—&gt;sql语句"></a>json格式—&gt;sql语句</h1><figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div><div class="line">10</div><div class="line">11</div><div class="line">12</div><div class="line">13</div><div class="line">14</div><div class="line">15</div><div class="line">16</div><div class="line">17</div></pre></td><td class="code"><pre><div class="line"><span class="comment">#!/usr/bin/env python</span></div><div class="line"><span class="comment">#-*- coding: utf8 -*-</span></div><div class="line"></div><div class="line"><span class="keyword">import</span> MySQLdb</div><div class="line"><span class="keyword">import</span> json</div><div class="line"></div><div class="line">datalist = []  　　　　　　　　　　　　　　　　　　　　<span class="comment"># python 列表</span></div><div class="line"><span class="keyword">with</span> open(<span class="string">'/data/scripts/logdb/json.txt'</span>,<span class="string">'r'</span>) <span class="keyword">as</span> f:</div><div class="line">    <span class="keyword">for</span> line <span class="keyword">in</span> f:　　　　　　　　　　　　　　　　　　　 <span class="comment"># 读取json文件中的行（也就是json的object）　　　</span></div><div class="line">        datalist.append(json.loads(line))　　　　　　　 <span class="comment"># 将json的object转成 Python的dict，追加到Python 列表中, 结果都是unicode格式：[&#123;&#125;,&#123;&#125;,&#123;&#125;,&#123;&#125;,&#123;&#125;] </span></div><div class="line"><span class="keyword">for</span> dict <span class="keyword">in</span> datalist:</div><div class="line">    <span class="keyword">print</span> dict　　　　　　　　　　　　　　　　　　　　 <span class="comment"># 打印显示 转换后的结果</span></div><div class="line"></div><div class="line"><span class="keyword">for</span> dict <span class="keyword">in</span> datalist:</div><div class="line">    dict[<span class="string">u'LOCAL'</span>] = dict[<span class="string">u'LOCAL'</span>].replace(<span class="string">'\r\n'</span>,<span class="string">'\\r\\n'</span>).replace(<span class="string">"'s"</span>,<span class="string">"\\'s"</span>)  <span class="comment"># 将字段中的特殊：回车换行以及's 转换，方便形成sql语句</span></div><div class="line">    sql = <span class="string">"insert into db1.s1 (mobile,NAME,LOCAL,CreateTime,id) values('%s','%s','%s','%s','%s');"</span> % (dict[<span class="string">u'mobile'</span>],dict[<span class="string">u'NAME'</span>],dict[<span class="string">u'LOCAL'</span>],str(dict[<span class="string">u'CreateTime'</span>]),dict[<span class="string">u'id'</span>])</div><div class="line">    <span class="keyword">print</span> sql</div></pre></td></tr></table></figure>
<p>本文转自：<a href="http://www.cnblogs.com/cuisi/p/6296224.html" target="_blank" rel="external">cnblogs</a></p>
]]></content>
    
    <summary type="html">
    
      &lt;blockquote&gt;
&lt;p&gt;今天碰到了一个比较奇葩的需求，将mysql表数据转换成json格式。我顺便研究了一下，写了个python。&lt;/p&gt;
&lt;/blockquote&gt;
&lt;hr&gt;
    
    </summary>
    
    
      <category term="databases" scheme="http://feiyang.li/tags/databases/"/>
    
      <category term="python" scheme="http://feiyang.li/tags/python/"/>
    
  </entry>
  
  <entry>
    <title>从零开始搭建Hadoop2.7.1的分布式集群</title>
    <link href="http://feiyang.li/2017/05/28/hadoop/index.html"/>
    <id>http://feiyang.li/2017/05/28/hadoop/index.html</id>
    <published>2017-05-28T11:57:31.000Z</published>
    <updated>2017-05-28T12:03:20.830Z</updated>
    
    <content type="html"><![CDATA[<script src="/assets/js/DPlayer.min.js"> </script><script src="/assets/js/APlayer.min.js"> </script><blockquote>
<p>Hadoop2.7.1，Hadoop的环境配置不是特别的复杂，但是确实有很多细节需要注意，不然会造成许多配置错误的情况。尽量保证一次配置正确防止反复修改。</p>
</blockquote>
<hr>
<a id="more"></a>
<h1 id="操作环境说明"><a href="#操作环境说明" class="headerlink" title="操作环境说明"></a>操作环境说明</h1><h2 id="操作系统"><a href="#操作系统" class="headerlink" title="操作系统"></a>操作系统</h2><ul>
<li>操作系统： window10</li>
<li>cpu： i7-6700k</li>
<li>内存：16g<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170528/171515618.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
</li>
</ul>
<h2 id="虚拟机版本：VMware12"><a href="#虚拟机版本：VMware12" class="headerlink" title="虚拟机版本：VMware12"></a>虚拟机版本：VMware12</h2><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170528/171637610.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<p>ssh-keygen -t rsa -C “mmmwhy@qq.com”</p>
<h1 id="材料准备"><a href="#材料准备" class="headerlink" title="材料准备"></a>材料准备</h1><ul>
<li>ubuntu-14.10-desktop-amd64.iso (Ubuntu 光盘映像)</li>
<li><a href="https://mirrors.tuna.tsinghua.edu.cn/apache/hadoop/common/hadoop-2.7.1/hadoop-2.7.1.tar.gz" target="_blank" rel="external">hadoop-2.7.1.tar.gz (Hadoop 环境包)</a><br>注意：<br>ubuntu14.04有自带的java，版本是1.7，输入<code>java -version</code>即可查看<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170528/172116248.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="mark" title="">
                </div>
                <div class="image-caption">mark</div>
            </figure>
如果没有安装java的话，网上找个安装java的文章即可，比如<a href="http://www.cnblogs.com/fnng/archive/2013/01/30/2883815.html" target="_blank" rel="external">这篇</a></li>
</ul>
<h1 id="搭建开始"><a href="#搭建开始" class="headerlink" title="搭建开始"></a>搭建开始</h1><h1 id="搭建开始-1"><a href="#搭建开始-1" class="headerlink" title="搭建开始"></a>搭建开始</h1><h2 id="Vmware-创建虚拟机"><a href="#Vmware-创建虚拟机" class="headerlink" title="Vmware 创建虚拟机"></a>Vmware 创建虚拟机</h2><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170528/172412587.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<p>根据Hadoop的调度规则，我们将使用VMware 12 加载 ubuntu….iso来创建三个Ubuntu 虚拟机。创建用典型安装即可。安装位置请选一个足够大的硬盘，预估需要50g。<strong>记住这个位置</strong><br><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170528/173230345.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure></p>
<p>这时我们先创建一个Ubuntu虚拟机，为什么这样稍后解释。</p>
<h1 id="SSH-安装以及配置"><a href="#SSH-安装以及配置" class="headerlink" title="SSH 安装以及配置"></a>SSH 安装以及配置</h1><h2 id="先配置ubuntu镜像源"><a href="#先配置ubuntu镜像源" class="headerlink" title="先配置ubuntu镜像源"></a>先配置ubuntu镜像源</h2><h3 id="打开软件源文件"><a href="#打开软件源文件" class="headerlink" title="打开软件源文件"></a>打开软件源文件</h3><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">sudo gedit /etc/apt/sources.list</div></pre></td></tr></table></figure>
<h3 id="将以下内容替换到源文件（注意在清华大学网站上选择合适自己系统的版本）"><a href="#将以下内容替换到源文件（注意在清华大学网站上选择合适自己系统的版本）" class="headerlink" title="将以下内容替换到源文件（注意在清华大学网站上选择合适自己系统的版本）"></a>将以下内容替换到源文件（注意在清华大学网站上选择合适自己系统的版本）</h3><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div><div class="line">10</div><div class="line">11</div><div class="line">12</div><div class="line">13</div></pre></td><td class="code"><pre><div class="line"># 默认注释了源码镜像以提高 apt update 速度，如有需要可自行取消注释</div><div class="line">deb https://mirrors.tuna.tsinghua.edu.cn/ubuntu/ trusty main restricted universe multiverse</div><div class="line"># deb-src https://mirrors.tuna.tsinghua.edu.cn/ubuntu/ trusty main restricted universe multiverse</div><div class="line">deb https://mirrors.tuna.tsinghua.edu.cn/ubuntu/ trusty-updates main restricted universe multiverse</div><div class="line"># deb-src https://mirrors.tuna.tsinghua.edu.cn/ubuntu/ trusty-updates main restricted universe multiverse</div><div class="line">deb https://mirrors.tuna.tsinghua.edu.cn/ubuntu/ trusty-backports main restricted universe multiverse</div><div class="line"># deb-src https://mirrors.tuna.tsinghua.edu.cn/ubuntu/ trusty-backports main restricted universe multiverse</div><div class="line">deb https://mirrors.tuna.tsinghua.edu.cn/ubuntu/ trusty-security main restricted universe multiverse</div><div class="line"># deb-src https://mirrors.tuna.tsinghua.edu.cn/ubuntu/ trusty-security main restricted universe multiverse</div><div class="line"></div><div class="line"># 预发布软件源，不建议启用</div><div class="line"># deb https://mirrors.tuna.tsinghua.edu.cn/ubuntu/ trusty-proposed main restricted universe multiverse</div><div class="line"># deb-src https://mirrors.tuna.tsinghua.edu.cn/ubuntu/ trusty-proposed main restricted universe multiverse</div></pre></td></tr></table></figure>
<h2 id="更新软件源"><a href="#更新软件源" class="headerlink" title="更新软件源"></a>更新软件源</h2><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">sudo apt-get update</div></pre></td></tr></table></figure>
<h2 id="安装ssh"><a href="#安装ssh" class="headerlink" title="安装ssh"></a>安装ssh</h2><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">sudo apt-get install openssh-server</div></pre></td></tr></table></figure>
<p>已有ssh或者安装成功了的输入命令<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">ps -e | grep ssh</div></pre></td></tr></table></figure></p>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170528/172819756.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<h2 id="验证SSH是否成功安装输入"><a href="#验证SSH是否成功安装输入" class="headerlink" title="验证SSH是否成功安装输入"></a>验证SSH是否成功安装输入</h2><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">ssh localhost</div></pre></td></tr></table></figure>
<p>出现以下提示说明安装成功<br><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170528/172945211.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure><br>注意，这个时候不是特权模式，特权模式反而会报错，因为我们设置的账户名其实不是<code>root</code></p>
<h2 id="生成密钥Pair"><a href="#生成密钥Pair" class="headerlink" title="生成密钥Pair"></a>生成密钥Pair</h2><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">ssh-keygen -t rsa</div></pre></td></tr></table></figure>
<p>输入后一直回车选择默认即可<br><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170528/173055646.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure><br>mater主机中输入命令复制一份公钥到home中<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">cp .ssh/id_rsa.pub ~/id_rsa_master.pub</div></pre></td></tr></table></figure></p>
<h1 id="复制出两个虚拟机"><a href="#复制出两个虚拟机" class="headerlink" title="复制出两个虚拟机"></a>复制出两个虚拟机</h1><p>到开始我们放置虚拟机的位置，复制两份放在旁边，然后用vm直接打开那两个文件夹。<br>这样可以生成三个一毛一样的虚拟机。<br><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170528/173515432.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure><br>顺便给他们把名字改成各自的角色。<br>这样做的好处是不需要将之前的步骤重复进行多次，不容易出错。</p>
<h1 id="配置-Hadoop-以下配置将仅仅在master主机上进行"><a href="#配置-Hadoop-以下配置将仅仅在master主机上进行" class="headerlink" title="配置 Hadoop (以下配置将仅仅在master主机上进行 )"></a>配置 Hadoop (<strong>以下配置将仅仅在master主机上进行 </strong>)</h1><ul>
<li>将解压缩好的hadoop-2.7.1文件夹拷贝到home根目录下面</li>
<li><p>在hadoop-2.7.1文件夹下创建文件，输入</p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div></pre></td><td class="code"><pre><div class="line">mkdir hadoop-2.7.1/tmp</div><div class="line">mkdir hadoop-2.7.1/hdfs</div><div class="line">mkdir hadoop-2.7.1/hdfs/name</div><div class="line">mkdir hadoop-2.7.1/hdfs/data</div></pre></td></tr></table></figure>
</li>
<li><p>输入命令查看ip地址 ，三个都要看，找个小本本记下。</p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">ifconfig -a</div></pre></td></tr></table></figure>
</li>
</ul>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170528/174451086.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<p>eg. 我所使用的IP地址<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div></pre></td><td class="code"><pre><div class="line">192.168.232.36 master</div><div class="line">192.168.232.54 slave1</div><div class="line">192.168.232.79 slave2</div></pre></td></tr></table></figure></p>
<ul>
<li>修改hosts<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">sudo gedit /etc/hosts</div></pre></td></tr></table></figure>
</li>
</ul>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170528/174509453.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<p>三个ip地址和对应的角色写在hosts内</p>
<ul>
<li>修改hostname<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">sudo gedit /etc/hostname</div></pre></td></tr></table></figure>
</li>
</ul>
<p>master 的改为 master<br>slave1 的改为 slave1<br>slave2 的改为 slave2</p>
<h2 id="修改环境变量"><a href="#修改环境变量" class="headerlink" title="修改环境变量"></a>修改环境变量</h2><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">cd ~/hadoop-2.7.1/</div></pre></td></tr></table></figure>
<h2 id="hadoop-env-sh"><a href="#hadoop-env-sh" class="headerlink" title="hadoop-env.sh"></a>hadoop-env.sh</h2><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">gedit etc/hadoop/hadoop-env.sh</div></pre></td></tr></table></figure>
<p>找到JAVA_HOME=… 一行修改为Java HOME的路径<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">export JAVA_HOME=/usr/lib/jvm/java-1.7.0-openjdk-amd64</div></pre></td></tr></table></figure></p>
<h2 id="core-site-xml"><a href="#core-site-xml" class="headerlink" title="core-site.xml"></a>core-site.xml</h2><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">gedit etc/hadoop/core-site.xml</div></pre></td></tr></table></figure>
<p>在configuration标签中添加<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div></pre></td><td class="code"><pre><div class="line">&lt;property&gt;</div><div class="line">    &lt;name&gt;fs.defaultFS&lt;/name&gt;</div><div class="line">    &lt;value&gt;hdfs://master:9000&lt;/value&gt;</div><div class="line">&lt;/property&gt;</div><div class="line"></div><div class="line">&lt;property&gt;</div><div class="line">    &lt;name&gt;hadoop.tmp.dir&lt;/name&gt;</div><div class="line">    &lt;value&gt;file:/home/mmmwhy/hadoop-2.7.1/tmp&lt;/value&gt;</div><div class="line">&lt;/property&gt;</div></pre></td></tr></table></figure></p>
<h2 id="mapred-site-xml"><a href="#mapred-site-xml" class="headerlink" title="mapred-site.xml"></a>mapred-site.xml</h2><p>创建并编辑<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div></pre></td><td class="code"><pre><div class="line">cp etc/hadoop/mapred-site.xml.template etc/hadoop/mapred-site.xml</div><div class="line">gedit etc/hadoop/mapred-site.xml</div></pre></td></tr></table></figure></p>
<p>在configuration标签中添加<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div></pre></td><td class="code"><pre><div class="line">&lt;property&gt;</div><div class="line">    &lt;name&gt;fs.defaultFS&lt;/name&gt;</div><div class="line">    &lt;value&gt;hdfs://master:9000&lt;/value&gt;</div><div class="line">&lt;/property&gt;</div><div class="line"></div><div class="line">&lt;property&gt;</div><div class="line">    &lt;name&gt;hadoop.tmp.dir&lt;/name&gt;</div><div class="line">    &lt;value&gt;file:/home/mmmwhy/hadoop-2.7.1/tmp&lt;/value&gt;</div><div class="line">&lt;/property&gt;</div></pre></td></tr></table></figure></p>
<h2 id="hdfs-site-xml"><a href="#hdfs-site-xml" class="headerlink" title="hdfs-site.xml"></a>hdfs-site.xml</h2><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">gedit etc/hadoop/hdfs-site.xml</div></pre></td></tr></table></figure>
<p>在configuration标签中添加<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div><div class="line">10</div><div class="line">11</div><div class="line">12</div><div class="line">13</div><div class="line">14</div><div class="line">15</div><div class="line">16</div></pre></td><td class="code"><pre><div class="line">&lt;property&gt;</div><div class="line">    &lt;name&gt;dfs.namenode.name.dir&lt;/name&gt;</div><div class="line">    &lt;value&gt;file:/home/mmmwhy/hadoop-2.7.1/hdfs/name&lt;/value&gt;</div><div class="line">&lt;/property&gt;</div><div class="line">&lt;property&gt;</div><div class="line">    &lt;name&gt;dfs.datanode.data.dir&lt;/name&gt;</div><div class="line">    &lt;value&gt;file:/home/mmmwhy/hadoop-2.7.1/hdfs/data&lt;/value&gt;</div><div class="line">&lt;/property&gt;</div><div class="line">&lt;property&gt;</div><div class="line">    &lt;name&gt;dfs.replication&lt;/name&gt;</div><div class="line">    &lt;value&gt;2&lt;/value&gt;</div><div class="line">&lt;/property&gt;</div><div class="line">&lt;property&gt;</div><div class="line">    &lt;name&gt;dfs.namenode.secondary.http-address&lt;/name&gt;</div><div class="line">    &lt;value&gt;master:9001&lt;/value&gt;</div><div class="line">&lt;/property&gt;</div></pre></td></tr></table></figure></p>
<h2 id="yarn-site-xml"><a href="#yarn-site-xml" class="headerlink" title="yarn-site.xml"></a>yarn-site.xml</h2><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">gedit etc/hadoop/yarn-site.xml</div></pre></td></tr></table></figure>
<p>在configuration标签中添加<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div><div class="line">10</div><div class="line">11</div><div class="line">12</div><div class="line">13</div><div class="line">14</div><div class="line">15</div><div class="line">16</div><div class="line">17</div><div class="line">18</div><div class="line">19</div><div class="line">20</div><div class="line">21</div><div class="line">22</div><div class="line">23</div><div class="line">24</div><div class="line">25</div><div class="line">26</div><div class="line">27</div><div class="line">28</div></pre></td><td class="code"><pre><div class="line">&lt;property&gt;</div><div class="line">    &lt;name&gt;yarn.nodemanager.aux-services&lt;/name&gt;</div><div class="line">    &lt;value&gt;mapreduce_shuffle&lt;/value&gt;</div><div class="line">&lt;/property&gt;</div><div class="line">&lt;property&gt;</div><div class="line">    &lt;name&gt;yarn.nodemanager.auxservices.mapreduce.shuffle.class&lt;/name&gt;</div><div class="line">    &lt;value&gt;org.apache.hadoop.mapred.ShuffleHandler&lt;/value&gt;</div><div class="line">&lt;/property&gt;</div><div class="line">&lt;property&gt;</div><div class="line">    &lt;name&gt;yarn.resourcemanager.address&lt;/name&gt;</div><div class="line">    &lt;value&gt;master:8032&lt;/value&gt;</div><div class="line">&lt;/property&gt;</div><div class="line">&lt;property&gt;</div><div class="line">    &lt;name&gt;yarn.resourcemanager.scheduler.address&lt;/name&gt;</div><div class="line">    &lt;value&gt;master:8030&lt;/value&gt;</div><div class="line">&lt;/property&gt;</div><div class="line">&lt;property&gt;</div><div class="line">    &lt;name&gt;yarn.resourcemanager.resource-tracker.address&lt;/name&gt;</div><div class="line">    &lt;value&gt;master:8031&lt;/value&gt;</div><div class="line">&lt;/property&gt;</div><div class="line">&lt;property&gt;</div><div class="line">    &lt;name&gt;yarn.resourcemanager.admin.address&lt;/name&gt;</div><div class="line">    &lt;value&gt;master:8033&lt;/value&gt;</div><div class="line">&lt;/property&gt;</div><div class="line">&lt;property&gt;</div><div class="line">    &lt;name&gt;yarn.resourcemanager.webapp.address&lt;/name&gt;</div><div class="line">    &lt;value&gt;master:8088&lt;/value&gt;</div><div class="line">&lt;/property&gt;</div></pre></td></tr></table></figure></p>
<h2 id="slaves文件"><a href="#slaves文件" class="headerlink" title="slaves文件"></a>slaves文件</h2><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">gedit etc/hadoop/slaves</div></pre></td></tr></table></figure>
<p>删除原有内容，根据配置修改，此处为<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div></pre></td><td class="code"><pre><div class="line">slave1</div><div class="line">slave2</div></pre></td></tr></table></figure></p>
<ul>
<li>分发配置好的hadoop文件夹到slave1, slave2<br>前提是设置好ssh<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div></pre></td><td class="code"><pre><div class="line">scp -r hadoop-2.7.1 mmmwhy@slave1:~/ </div><div class="line">scp -r hadoop-2.7.1 mmmwhy@slave2:~/</div></pre></td></tr></table></figure>
</li>
</ul>
<p>应该出现一大堆看起来非常厉害的东西。</p>
<ul>
<li><p>格式化hdfs<br>进入hadoop home目录</p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">bin/hdfs namenode -format</div></pre></td></tr></table></figure>
</li>
<li><p>启动集群</p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">sbin/start-all.sh</div></pre></td></tr></table></figure>
</li>
</ul>
<p>启动后分别在master, slave下输入jps查看进程<br><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170528/175313622.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure><br>如上则说明启动成功</p>
<h1 id="集群测试"><a href="#集群测试" class="headerlink" title="集群测试"></a>集群测试</h1><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div></pre></td><td class="code"><pre><div class="line">echo hellecho hello world &gt;&gt; in1.txt hadoop &gt;&gt; in2.txt</div><div class="line">cd ~/mmmwhy/hadoop-2.7.1/</div><div class="line">bin/hadoop fs -mkdir /usr</div><div class="line">bin/hadoop fs -mkdir /usr/input</div><div class="line">bin/hadoop fs -put in1.txt /usr/input</div><div class="line">bin/hadoop fs -put in2.txt /usr/input</div><div class="line">#执行Hadoop内置example测试程序</div><div class="line">bin/hadoop jar ~/hadoop-2.7.1/share/hadoop/mapreduce/hadoop-mapreduce-examples-2.7.1.jar wordcount /usr/input /usr/output</div></pre></td></tr></table></figure>
<p>然后进入 <a href="http://localhost:50070/explorer.html#/usr/input" target="_blank" rel="external">http://localhost:50070/explorer.html#/usr/input</a><br>就可以看到<br><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170528/194607098.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure></p>
<h1 id="运行wordcount测试"><a href="#运行wordcount测试" class="headerlink" title="运行wordcount测试"></a>运行wordcount测试</h1><p>进入hadoop home目录<br>在hdfs（分布式文件系统）中创建一个名为input的文件夹<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">bin/hadoop fs -mkdir /input</div></pre></td></tr></table></figure></p>
<p>查看文件是否被创建<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">bin/hadoop fs -ls /</div></pre></td></tr></table></figure></p>
<p>hadoop home 下创建一个inputfile 文件夹，并在inputfile里创建两个文件<br><strong>in1.txt</strong><br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">Hello world hello hadoop</div></pre></td></tr></table></figure></p>
<p><strong>in2.txt</strong><br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">Hello Hadoop hello whatever</div></pre></td></tr></table></figure></p>
<p>上传两个文件进input<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">bin/hadoop fs -put inputfile/*.txt /input</div></pre></td></tr></table></figure></p>
<p>查看输入文件是否传入<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">bin/hadoop fs -ls /input</div></pre></td></tr></table></figure></p>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170528/195114806.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<p>用hadoop jar命令运行Hadoop自带的wordcount<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">bin/hadoop jar share/hadoop/mapreduce/hadoop-mapreduce-examples-2.7.1.jar wordcount /input /output</div></pre></td></tr></table></figure></p>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170528/195202978.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<p>程序开始运行，成功后查看输出文件夹<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div></pre></td><td class="code"><pre><div class="line">bin/hadoop fs -ls /output</div><div class="line">bin/hadoop fs -cat /output/part-r-00000</div></pre></td></tr></table></figure></p>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170528/195431835.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<p>至此hadoop分布式集群配置完成！本文主体内容转自<a href="http://blog.csdn.net/zl007700/article/details/50533675" target="_blank" rel="external">CSDN</a>，原作者写的非常好，但是有一些小问题，做到一半的时候花式报错，因此自己写了一份，修正了其中一些问题。</p>
]]></content>
    
    <summary type="html">
    
      &lt;blockquote&gt;
&lt;p&gt;Hadoop2.7.1，Hadoop的环境配置不是特别的复杂，但是确实有很多细节需要注意，不然会造成许多配置错误的情况。尽量保证一次配置正确防止反复修改。&lt;/p&gt;
&lt;/blockquote&gt;
&lt;hr&gt;
    
    </summary>
    
    
      <category term="教程" scheme="http://feiyang.li/tags/%E6%95%99%E7%A8%8B/"/>
    
      <category term="linux" scheme="http://feiyang.li/tags/linux/"/>
    
      <category term="hadoop" scheme="http://feiyang.li/tags/hadoop/"/>
    
  </entry>
  
  <entry>
    <title>算法实验(一)</title>
    <link href="http://feiyang.li/2017/05/24/algorithm1/index.html"/>
    <id>http://feiyang.li/2017/05/24/algorithm1/index.html</id>
    <published>2017-05-24T10:48:48.000Z</published>
    <updated>2017-05-31T01:40:39.589Z</updated>
    
    <content type="html"><![CDATA[<script src="/assets/js/DPlayer.min.js"> </script><script src="/assets/js/APlayer.min.js"> </script><blockquote>
<p>快速排序与背包问题相关算法分析</p>
</blockquote>
<hr>
<a id="more"></a>
<h1 id="快速排序"><a href="#快速排序" class="headerlink" title="快速排序"></a>快速排序</h1><p><a href="http://feiyang.li/2017/03/09/Python-quick-order/index.html">相关文章</a></p>
<h2 id="普通快速排序"><a href="#普通快速排序" class="headerlink" title="普通快速排序"></a>普通快速排序</h2><h3 id="算法效率："><a href="#算法效率：" class="headerlink" title="算法效率："></a>算法效率：</h3><ul>
<li>最坏情况下分析：当输入序列是正序或者是反序的时候，效率最坏，这时效率是$Θ(n^2)$</li>
<li>最优情况下分析：其他情况下的算法效率趋向于$Θ（nlgn）$<h3 id="伪代码"><a href="#伪代码" class="headerlink" title="伪代码"></a>伪代码</h3>对整个数组进行递归排序：<figure class="highlight cpp"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div></pre></td><td class="code"><pre><div class="line">QUICKSORT(A,p,r)</div><div class="line"><span class="keyword">if</span> p &lt; r:</div><div class="line">    q = PARTITION(A,p,r);</div><div class="line">    QUICKSORT(A,p,q<span class="number">-1</span>);</div><div class="line">    QUICKSORT(A,q+<span class="number">1</span>,r);</div></pre></td></tr></table></figure>
</li>
</ul>
<p>快速排序算法的关键是PARTITION过程，它对A[p..r]进行就地重排：<br><figure class="highlight cpp"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div></pre></td><td class="code"><pre><div class="line">PARTITION(A,p,r)</div><div class="line">x = A[r]</div><div class="line">i = p<span class="number">-1</span></div><div class="line"><span class="keyword">for</span> j = p to r<span class="number">-1</span><span class="comment">// A[j]是待比较的元素</span></div><div class="line">  <span class="keyword">if</span> A[j] ≤ r<span class="comment">// 若A[j]比主元小</span></div><div class="line">    i = i + <span class="number">1</span> <span class="comment">// i往后一位（符合，不用替换；不符合i就停在这个地方，等着j过来叫唤）</span></div><div class="line">    exchange A[i] with A[j]<span class="comment">//倘若A[j]比主元小，其实是不发生交换的。</span></div><div class="line">exchange A[i+<span class="number">1</span>] with A[r]</div><div class="line"><span class="keyword">return</span> i + <span class="number">1</span></div></pre></td></tr></table></figure></p>
<h3 id="python代码"><a href="#python代码" class="headerlink" title="python代码"></a>python代码</h3><figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div><div class="line">10</div><div class="line">11</div><div class="line">12</div><div class="line">13</div><div class="line">14</div><div class="line">15</div><div class="line">16</div><div class="line">17</div><div class="line">18</div><div class="line">19</div><div class="line">20</div><div class="line">21</div><div class="line">22</div><div class="line">23</div><div class="line">24</div><div class="line">25</div><div class="line">26</div><div class="line">27</div><div class="line">28</div><div class="line">29</div><div class="line">30</div><div class="line">31</div><div class="line">32</div><div class="line">33</div><div class="line">34</div></pre></td><td class="code"><pre><div class="line"><span class="comment"># coding=utf-8</span></div><div class="line"><span class="keyword">import</span> datetime</div><div class="line"><span class="keyword">import</span> sys   </div><div class="line">sys.setrecursionlimit(<span class="number">1000000</span>)</div><div class="line"><span class="function"><span class="keyword">def</span> <span class="title">SET_NUMBER</span><span class="params">()</span>:</span></div><div class="line">    l = []</div><div class="line">    <span class="keyword">for</span> i <span class="keyword">in</span> range(<span class="number">0</span>,len(range(<span class="number">1</span>,<span class="number">2000</span>,<span class="number">2</span>)[::<span class="number">-1</span>])<span class="number">-1</span>):</div><div class="line">        l.append(range(<span class="number">1</span>,<span class="number">2000</span>,<span class="number">2</span>)[::<span class="number">-1</span>][i])</div><div class="line">    <span class="keyword">return</span> l</div><div class="line">        </div><div class="line"><span class="function"><span class="keyword">def</span> <span class="title">PARTITION</span><span class="params">(A,p,r)</span>:</span></div><div class="line">    x = A[r]</div><div class="line">    i = p - <span class="number">1</span></div><div class="line">    <span class="keyword">for</span> j <span class="keyword">in</span> range(p,r):</div><div class="line">        <span class="keyword">if</span> A[j]&lt;=x:</div><div class="line">            i = i + <span class="number">1</span></div><div class="line">            A[i],A[j]=A[j],A[i]</div><div class="line">    A[i+<span class="number">1</span>],A[r] = A[r],A[i+<span class="number">1</span>]</div><div class="line">    <span class="keyword">return</span> i+<span class="number">1</span></div><div class="line"></div><div class="line"></div><div class="line"><span class="function"><span class="keyword">def</span> <span class="title">QUICKSORT</span><span class="params">(A,p,r)</span>:</span></div><div class="line">    <span class="keyword">if</span> p &gt;= r:</div><div class="line">        <span class="keyword">return</span></div><div class="line">    q = PARTITION(A,p,r)</div><div class="line">    QUICKSORT(A,p,q<span class="number">-1</span>)</div><div class="line">    QUICKSORT(A,q+<span class="number">1</span>,r)</div><div class="line">    </div><div class="line"><span class="keyword">if</span> __name__ == <span class="string">'__main__'</span>:</div><div class="line">    l = SET_NUMBER()</div><div class="line">    time1 = datetime.datetime.now()</div><div class="line">    QUICKSORT(l,<span class="number">0</span>,len(l)<span class="number">-1</span>)</div><div class="line">    time2 = datetime.datetime.now()</div><div class="line">    print(time2-time1)</div></pre></td></tr></table></figure>
<h2 id="随机快速排序"><a href="#随机快速排序" class="headerlink" title="随机快速排序"></a>随机快速排序</h2><h3 id="算法效率"><a href="#算法效率" class="headerlink" title="算法效率"></a>算法效率</h3><p>我们已经知道，若输入本身已被排序，那么对于快排来说就糟了。那么如何避免这样的情况？一种方法时随机排列序列中的元素；另一种方法时随机地选择主元（pivot）。这便是随机化快速排序的思想，这种快排的好处是：其运行时间不依赖于输入序列的顺序。</p>
<p>经分析， 随机化快排的算法效率是$Θ(nlgn)$</p>
<h3 id="伪代码-1"><a href="#伪代码-1" class="headerlink" title="伪代码"></a>伪代码</h3><p>下面是分化（Partition）<br><figure class="highlight cpp"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div></pre></td><td class="code"><pre><div class="line">Random_Partition(<span class="built_in">vector</span>&lt;T&gt; &amp;A,<span class="keyword">int</span> p,<span class="keyword">int</span> q)</div><div class="line">    <span class="keyword">int</span> i=rand()%(q-p)+p;  <span class="comment">//此行与快排不同、加入随机数参数器</span></div><div class="line">    Swap(A[i],A[p]);         <span class="comment">//此行与快排不同、随机选取主元</span></div><div class="line">    <span class="keyword">return</span> Partition(A,p,q);<span class="comment">//此次与快速排序一样</span></div></pre></td></tr></table></figure></p>
<p>随机化快速排序<br><figure class="highlight cpp"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div></pre></td><td class="code"><pre><div class="line">Random_Quick_Sort(<span class="built_in">vector</span>&lt;T&gt; &amp;A,<span class="keyword">int</span> p,<span class="keyword">int</span> q)</div><div class="line">    <span class="keyword">if</span> (p&lt;q):</div><div class="line">        <span class="keyword">int</span> i=Random_Partition(A,p,q);</div><div class="line">        Random_Quick_Sort(A,p,i<span class="number">-1</span>);</div><div class="line">        Random_Quick_Sort(A,i+<span class="number">1</span>,q);</div></pre></td></tr></table></figure></p>
<h3 id="python代码-1"><a href="#python代码-1" class="headerlink" title="python代码"></a>python代码</h3><figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div><div class="line">10</div><div class="line">11</div><div class="line">12</div><div class="line">13</div><div class="line">14</div><div class="line">15</div><div class="line">16</div><div class="line">17</div><div class="line">18</div><div class="line">19</div><div class="line">20</div><div class="line">21</div><div class="line">22</div><div class="line">23</div><div class="line">24</div><div class="line">25</div><div class="line">26</div><div class="line">27</div><div class="line">28</div><div class="line">29</div><div class="line">30</div><div class="line">31</div><div class="line">32</div><div class="line">33</div><div class="line">34</div><div class="line">35</div><div class="line">36</div><div class="line">37</div><div class="line">38</div></pre></td><td class="code"><pre><div class="line"><span class="comment"># coding=utf-8</span></div><div class="line"><span class="keyword">import</span> datetime</div><div class="line"><span class="keyword">import</span> random</div><div class="line"><span class="function"><span class="keyword">def</span> <span class="title">SET_NUMBER</span><span class="params">()</span>:</span></div><div class="line">    l = []</div><div class="line">    <span class="keyword">for</span> i <span class="keyword">in</span> range(<span class="number">0</span>,len(range(<span class="number">1</span>,<span class="number">2000</span>,<span class="number">2</span>)[::<span class="number">-1</span>])<span class="number">-1</span>):</div><div class="line">        l.append(range(<span class="number">1</span>,<span class="number">2000</span>,<span class="number">2</span>)[::<span class="number">-1</span>][i])</div><div class="line">    <span class="keyword">return</span> l</div><div class="line"></div><div class="line"></div><div class="line"><span class="function"><span class="keyword">def</span> <span class="title">PARTITION</span><span class="params">(A,p,r)</span>:</span></div><div class="line">    x = A[r]</div><div class="line">    i = p - <span class="number">1</span></div><div class="line">    <span class="keyword">for</span> j <span class="keyword">in</span> range(p,r):</div><div class="line">        <span class="keyword">if</span> A[j]&lt;=x:</div><div class="line">            i = i + <span class="number">1</span></div><div class="line">            A[i],A[j]=A[j],A[i]</div><div class="line">    A[i+<span class="number">1</span>],A[r] = A[r],A[i+<span class="number">1</span>]</div><div class="line">    <span class="keyword">return</span> i+<span class="number">1</span></div><div class="line"></div><div class="line"><span class="function"><span class="keyword">def</span> <span class="title">Random_Partition</span><span class="params">(A,p,r)</span>:</span></div><div class="line">    i = random.randint(p, r) </div><div class="line">    A[i],A[p]=A[p],A[i]</div><div class="line">    <span class="keyword">return</span> PARTITION(A,p,r)</div><div class="line"></div><div class="line"><span class="function"><span class="keyword">def</span> <span class="title">QUICKSORT</span><span class="params">(A,p,r)</span>:</span></div><div class="line">    <span class="keyword">if</span> p &gt;= r:</div><div class="line">        <span class="keyword">return</span></div><div class="line">    q = Random_Partition(A,p,r)</div><div class="line">    QUICKSORT(A,p,q<span class="number">-1</span>)</div><div class="line">    QUICKSORT(A,q+<span class="number">1</span>,r)</div><div class="line">    </div><div class="line"><span class="keyword">if</span> __name__ == <span class="string">'__main__'</span>:</div><div class="line">    l = SET_NUMBER()</div><div class="line">    time1 = datetime.datetime.now()</div><div class="line">    QUICKSORT(l,<span class="number">0</span>,len(l)<span class="number">-1</span>)</div><div class="line">    time2 = datetime.datetime.now()</div><div class="line">    print(time2-time1)</div></pre></td></tr></table></figure>
<h2 id="时间效率"><a href="#时间效率" class="headerlink" title="时间效率"></a>时间效率</h2><ul>
<li>对于倒序数组<code>range(1,2000,2)[::-1]</code>(1000个数字)<ul>
<li>使用正常快速排序时间为：0:00:00.082566</li>
<li>使用随机快速排序时间为：0:00:00.005815</li>
</ul>
</li>
<li>对于随机数组<code>random.randint(0,9999)</code>(重复1000次)<ul>
<li>使用正常快速排序时间为：0:00:00.002974</li>
<li>使用随机快速排序时间为：0:00:00.004901</li>
</ul>
</li>
</ul>
<h2 id="分析性能差异"><a href="#分析性能差异" class="headerlink" title="分析性能差异"></a>分析性能差异</h2><p>快速排序的平均时间复杂度为$O(nlogn)$，最坏时间时间可达到$O(n^2)$，最坏情况是当要排序的数列基本有序的时候。根据快速排序的工作原理我们知道，选取第一个或者最后一个元素作为主元，快速排序依次将数列分成两个数列，然后递归将分开的数列进行排序。</p>
<p>当把数列平均分成两个等长的数列时，效率是最高的，如果相差太大，效率就会降低。 </p>
<p>我们通过使用随机化的快速排序随机的从数列中选取一个元素与第一个，或者最后一个元素交换，这样就会使得数列有序的概率降低。</p>
<p>所以随机快速排序平均速度是比快速排序要快的，但是因为随机排序需要有设置随机数+交换元素的时间，因此总时间上不如普通快速排序。</p>
<h1 id="背包问题"><a href="#背包问题" class="headerlink" title="背包问题"></a>背包问题</h1><p><a href="http://feiyang.li/2017/03/24/dynamic-programming/index.html">相关文章</a></p>
<h2 id="动态规划知识复习"><a href="#动态规划知识复习" class="headerlink" title="动态规划知识复习"></a>动态规划知识复习</h2><h3 id="基本步骤"><a href="#基本步骤" class="headerlink" title="基本步骤"></a>基本步骤</h3><ul>
<li>找出最优解的性质，并刻画其结构特征</li>
<li>递归地定义最优解的值</li>
<li>以自底而上的方式计算出最优值</li>
<li>根据计算最优值时得到的信息，构造一个最优解。<h3 id="若给定问题具有以下性质："><a href="#若给定问题具有以下性质：" class="headerlink" title="若给定问题具有以下性质："></a>若给定问题具有以下性质：</h3></li>
<li><strong>最优子结构</strong>：如果一个问题的最优解中包含了其子问题的最优解，就说该问题具有最优子结构。当一个问题具有最优子结构时，提示我们动态规划法可能会适用，但是此时贪心策略可能也是适用的。</li>
<li><strong>重叠子问题</strong>：指用来解原问题的递归算法可反复地解同样的子问题，而不是总在产生新的子问题。即当一个递归算法不断地调用同一个问题时，就说该问题包含重叠子问题。此时若用分治法递归求解，则每次遇到子问题都会视为新问题，会极大地降低算法的效率，而动态规划法总是充分利用重叠子问题，对于每个子问题仅计算一次，把解保存在一个在需要时就可以查看的表中，而每次查表的时间为常数。</li>
</ul>
<h2 id="0-1背包问题"><a href="#0-1背包问题" class="headerlink" title="0-1背包问题"></a>0-1背包问题</h2><h3 id="定义问题"><a href="#定义问题" class="headerlink" title="定义问题"></a>定义问题</h3><p>问题：有n个物品，第i个物品价值为$v_i$，重量为$w_i$，其中$v_i$和$w_i$均为非负数，背包的容量为W，W为非负数。现需要考虑如何选择装入背包的物品，使装入背包的物品总价值最大。</p>
<p>该问题以形式化描述如下：</p>
<ul>
<li>目标函数为 ：<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://img.my.csdn.net/uploads/201303/24/1364096225_6763.jpg?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure></li>
<li>约束条件为：<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://img.my.csdn.net/uploads/201303/24/1364096448_8646.jpg?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
满足约束条件的任一集合（x1，x2，…，xn）是问题的一个可行解，问题的目标是要求问题的一个最优解。考虑一个实例，假设n=5，W=17， 每个物品的价值和重量如下表所示。可将物品1，2和5装入背包，背包未满，获得价值22，此时问题解为你（1，1，0，0，1）。也可以将物品4和5装入背包，背包装满，获得价值24，此时解为（0，0，0，1，1）。<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://img.my.csdn.net/uploads/201303/24/1364096951_8384.jpg?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
</li>
</ul>
<h3 id="按步骤求解问题"><a href="#按步骤求解问题" class="headerlink" title="按步骤求解问题"></a>按步骤求解问题</h3><ul>
<li>找出最优解的性质，并刻画其结构特征<br>可以将背包问题的求解过程看作是进行一系列的决策过程，即决定哪些物品应该放入背包，哪些物品不放入背包。如果一个问题的最优解包含了物品n，即$x_n=1$，那么其余$x_1，x<em>2，…，x</em>(n-1)$一定构成子问题1，2，…，n-1在容量W-wn时的最优解。如果这个最优解不包含物品n，即$x_n$=0，那么其余$x_1，x<em>2，…，x</em>(n-1)$一定构成子问题1，2，…，n-1在容量W时的最优解。</li>
<li>递归定义最优解的值<br>根据上述分析的最优解的结构递归地定义问题最优解。设c[i,w]表示背包剩余容量为w时，i个物品导致的最优解的总价值。<br>分为三种情况：<ul>
<li>i或者w等于0，没有空间也没有物品的时候；</li>
<li>第i的物品的重量已经超过当前空间时；</li>
<li>在增加第i个物品和不增加第i个物品之间做出艰难的抉择；<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://img.my.csdn.net/uploads/201303/24/1364126887_3091.jpg?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure></li>
</ul>
</li>
<li>以自底而上的方式计算出最优值<br>伪代码：<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div></pre></td><td class="code"><pre><div class="line">BOTTOM-UP-CUT-ROD(p,n)</div><div class="line">let  r[0..n] be a new array</div><div class="line">r[0]= 0</div><div class="line">for  j = 1 to n</div><div class="line">    q= -∞</div><div class="line">    for  i = 1 to j</div><div class="line">        q = max(q, p[i]+r[j-i])</div><div class="line">    r[j]= q</div><div class="line">return  r[n]</div></pre></td></tr></table></figure>
</li>
</ul>
<h3 id="python代码-2"><a href="#python代码-2" class="headerlink" title="python代码"></a>python代码</h3><figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div><div class="line">10</div><div class="line">11</div><div class="line">12</div><div class="line">13</div><div class="line">14</div><div class="line">15</div><div class="line">16</div><div class="line">17</div><div class="line">18</div><div class="line">19</div><div class="line">20</div><div class="line">21</div><div class="line">22</div><div class="line">23</div><div class="line">24</div><div class="line">25</div><div class="line">26</div><div class="line">27</div><div class="line">28</div><div class="line">29</div><div class="line">30</div><div class="line">31</div></pre></td><td class="code"><pre><div class="line"><span class="comment">#coding:utf-8  </span></div><div class="line"><span class="function"><span class="keyword">def</span> <span class="title">bag</span><span class="params">(n,m,w,v)</span>:</span>  </div><div class="line">    res = [[<span class="number">0</span> <span class="keyword">for</span> j <span class="keyword">in</span> range(m+<span class="number">1</span>)]<span class="keyword">for</span> i <span class="keyword">in</span> range(n+<span class="number">1</span>)] <span class="comment">#n+1 行，m+1列 值为0的矩阵  </span></div><div class="line">    <span class="keyword">for</span> i <span class="keyword">in</span> range(<span class="number">1</span>,n+<span class="number">1</span>):  </div><div class="line">        <span class="keyword">for</span> j <span class="keyword">in</span> range(<span class="number">1</span>,m+<span class="number">1</span>):  </div><div class="line">            res[i][j] = res[i<span class="number">-1</span>][j] <span class="comment">#0-&gt;res[0][1]-&gt;res[1][1]  </span></div><div class="line">            <span class="keyword">if</span> j&gt;=w[i<span class="number">-1</span>] <span class="keyword">and</span> res[i][j]&lt;res[i<span class="number">-1</span>][j-w[i<span class="number">-1</span>]]+v[i<span class="number">-1</span>]:<span class="comment">#如果res[i-1][j]小于res[i-1][j-w[i-1]]+v[i-1]，那么res[i][j]就等于res[i-1][j]，否则就等于res[i-1][j-w[i-1]]+v[i-1]  </span></div><div class="line">                res[i][j] = res[i<span class="number">-1</span>][j-w[i<span class="number">-1</span>]]+v[i<span class="number">-1</span>]  </div><div class="line">    <span class="keyword">return</span> res  </div><div class="line">  </div><div class="line"><span class="function"><span class="keyword">def</span> <span class="title">show</span><span class="params">(n,m,w,res)</span>:</span>  </div><div class="line">    <span class="keyword">print</span> (<span class="string">u"最大值为%d"</span>%res[n][m]) </div><div class="line">    x = [<span class="keyword">False</span> <span class="keyword">for</span> i <span class="keyword">in</span> range(n)]  </div><div class="line">    j = m  </div><div class="line">    <span class="keyword">for</span>  i <span class="keyword">in</span> range(n,<span class="number">0</span>,<span class="number">-1</span>):  </div><div class="line">        <span class="keyword">if</span> res[i][j]!=res[i<span class="number">-1</span>][j]:  </div><div class="line">            x[i<span class="number">-1</span>] = <span class="keyword">True</span>  </div><div class="line">            j-=w[i<span class="number">-1</span>]  </div><div class="line">    <span class="keyword">print</span> (<span class="string">u"选择的物品为"</span>)  </div><div class="line">    <span class="keyword">for</span> i <span class="keyword">in</span> range(n):  </div><div class="line">        <span class="keyword">if</span> x[i]:  </div><div class="line">            print(<span class="string">u"第%d个"</span>%(i+<span class="number">1</span>))  </div><div class="line">  </div><div class="line"><span class="keyword">if</span> __name__ == <span class="string">"__main__"</span>:  </div><div class="line">    <span class="comment">#n种物品，承重量为m，w物品的重量，v 物品的价值  </span></div><div class="line">    n= <span class="number">5</span>  </div><div class="line">    m = <span class="number">17</span>  </div><div class="line">    w = [<span class="number">3</span>,<span class="number">4</span>,<span class="number">7</span>,<span class="number">8</span>,<span class="number">9</span>] </div><div class="line">    v = [<span class="number">4</span>,<span class="number">5</span>,<span class="number">10</span>,<span class="number">11</span>,<span class="number">13</span>]  </div><div class="line">    res = bag(n,m,w,v)  </div><div class="line">    show(n,m,w,res)</div></pre></td></tr></table></figure>
<h2 id="部分背包问题"><a href="#部分背包问题" class="headerlink" title="部分背包问题"></a>部分背包问题</h2><p>题目要求使用贪心算法</p>
<ul>
<li>部分背包问题与0-1背包问题的差别在于：可以带走一部分。即，部分背包问题可带走的物品 是可以 无限细分的。</li>
<li>贪心策略是：总是优先选择单位重量下价值最大的物品<h3 id="具体题目"><a href="#具体题目" class="headerlink" title="具体题目"></a>具体题目</h3>假设背包可容纳50Kg的重量，物品信息如下：</li>
</ul>
<table>
<thead>
<tr>
<th>物品 i</th>
<th style="text-align:left">重量(Kg)</th>
<th>价值</th>
<th style="text-align:left">单位重量的价值</th>
</tr>
</thead>
<tbody>
<tr>
<td>1</td>
<td style="text-align:left">10</td>
<td>60</td>
<td style="text-align:left">6</td>
</tr>
<tr>
<td>2</td>
<td style="text-align:left">20</td>
<td>100</td>
<td style="text-align:left">5</td>
</tr>
<tr>
<td>3</td>
<td style="text-align:left">30</td>
<td>120</td>
<td style="text-align:left">4</td>
</tr>
</tbody>
</table>
<h3 id="伪代码-2"><a href="#伪代码-2" class="headerlink" title="伪代码"></a>伪代码</h3><p>这个看起来就很简单，先装单价最高的1，剩下的地方装2，再剩下的地方装3<br>我也不知道怎么写伪代码，直接写python吧</p>
<h3 id="C代码"><a href="#C代码" class="headerlink" title="C代码"></a>C代码</h3><figure class="highlight cpp"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div><div class="line">10</div><div class="line">11</div><div class="line">12</div><div class="line">13</div><div class="line">14</div><div class="line">15</div><div class="line">16</div><div class="line">17</div><div class="line">18</div><div class="line">19</div><div class="line">20</div><div class="line">21</div><div class="line">22</div><div class="line">23</div><div class="line">24</div><div class="line">25</div><div class="line">26</div><div class="line">27</div><div class="line">28</div><div class="line">29</div><div class="line">30</div><div class="line">31</div><div class="line">32</div><div class="line">33</div><div class="line">34</div><div class="line">35</div><div class="line">36</div><div class="line">37</div><div class="line">38</div><div class="line">39</div><div class="line">40</div><div class="line">41</div><div class="line">42</div><div class="line">43</div><div class="line">44</div><div class="line">45</div><div class="line">46</div><div class="line">47</div><div class="line">48</div><div class="line">49</div><div class="line">50</div><div class="line">51</div><div class="line">52</div><div class="line">53</div><div class="line">54</div><div class="line">55</div><div class="line">56</div></pre></td><td class="code"><pre><div class="line"><span class="meta">#<span class="meta-keyword">include</span> <span class="meta-string">&lt;stdio.h&gt;</span>  </span></div><div class="line"><span class="meta">#<span class="meta-keyword">include</span> <span class="meta-string">&lt;algorithm&gt;</span>  </span></div><div class="line"><span class="keyword">using</span> <span class="keyword">namespace</span> <span class="built_in">std</span>;  </div><div class="line">  </div><div class="line">  </div><div class="line"><span class="comment">//物品结构体，包含两个属性，w表示重量，v表示价值  </span></div><div class="line"><span class="keyword">struct</span> Good&#123;  </div><div class="line">    <span class="keyword">int</span> id;  </div><div class="line">    <span class="keyword">double</span> w;  </div><div class="line">    <span class="keyword">double</span> v;  </div><div class="line">&#125;;  </div><div class="line">  </div><div class="line"><span class="comment">//排序比较函数，以物品的价值/重量比值降序排序  </span></div><div class="line"><span class="function"><span class="keyword">bool</span> <span class="title">cmp</span><span class="params">(Good a, Good b)</span></span>&#123;  </div><div class="line">    <span class="keyword">if</span> (a.v/a.w &gt; b.v/b.w) &#123;  </div><div class="line">        <span class="keyword">return</span> <span class="literal">true</span>;  </div><div class="line">    &#125;  </div><div class="line">    <span class="keyword">return</span> <span class="literal">false</span>;  </div><div class="line">&#125;  </div><div class="line">  </div><div class="line"><span class="comment">//结构体数组,所有物品信息  </span></div><div class="line"><span class="keyword">struct</span> Good goods[] = &#123;&#123;<span class="number">1</span>,<span class="number">10</span>,<span class="number">12</span>&#125;,&#123;<span class="number">2</span>,<span class="number">9</span>,<span class="number">9</span>&#125;,&#123;<span class="number">3</span>,<span class="number">11</span>,<span class="number">7</span>&#125;,&#123;<span class="number">4</span>,<span class="number">12</span>,<span class="number">9</span>&#125;,&#123;<span class="number">5</span>,<span class="number">6</span>,<span class="number">8</span>&#125;,&#123;<span class="number">6</span>,<span class="number">3</span>,<span class="number">4</span>&#125;&#125;;  </div><div class="line">  </div><div class="line">  </div><div class="line"><span class="comment">//背包总重量  </span></div><div class="line"><span class="keyword">double</span> totalW = <span class="number">30</span>;  </div><div class="line">  </div><div class="line"><span class="function"><span class="keyword">int</span> <span class="title">main</span><span class="params">()</span>  </span></div><div class="line">&#123;  </div><div class="line">    <span class="comment">//以物品的价值/重量比值降序排序  </span></div><div class="line">    sort(goods,goods+<span class="number">6</span>,cmp);  </div><div class="line">    <span class="keyword">double</span> leftW = totalW;  </div><div class="line">    <span class="keyword">int</span> totalV = <span class="number">0</span>;  </div><div class="line">      </div><div class="line">    <span class="comment">//遍历排好序的物品数组  </span></div><div class="line">    <span class="keyword">for</span> (<span class="keyword">int</span> i=<span class="number">0</span>; i&lt;<span class="number">6</span>; i++) &#123;  </div><div class="line">        <span class="comment">//如果当前背包所能承受的重量大于i物品的总量  </span></div><div class="line">        <span class="comment">//那么把i物品全部放进去  </span></div><div class="line">        <span class="keyword">if</span> (leftW &gt;= goods[i].w) &#123;  </div><div class="line">            leftW -= goods[i].w;  </div><div class="line">            totalV += goods[i].v;  </div><div class="line">            <span class="built_in">printf</span>(<span class="string">"choose good[id = %d], %.1f weight,make %.1f value\n"</span>,goods[i].id,goods[i].w,goods[i].v);  </div><div class="line">              </div><div class="line">        <span class="comment">//如果不能，那么取当前背包所能承受重量的相应数量物品  </span></div><div class="line">        <span class="comment">//当然价值也得按照比例来  </span></div><div class="line">        &#125;<span class="keyword">else</span> &#123;  </div><div class="line">            totalV += leftW/goods[i].w * goods[i].v;  </div><div class="line">            leftW = <span class="number">0</span>;  </div><div class="line">            <span class="built_in">printf</span>(<span class="string">"choose good[id = %d], %.1f weight,make %.1f value\n"</span>,goods[i].id,leftW,leftW/goods[i].w * goods[i].v);  </div><div class="line">            <span class="keyword">break</span>;  </div><div class="line">        &#125;  </div><div class="line">    &#125;  </div><div class="line">      </div><div class="line">    <span class="built_in">printf</span>(<span class="string">"max total value:%d\n"</span>,totalV);  </div><div class="line">    <span class="keyword">return</span> <span class="number">0</span>;  </div><div class="line">&#125;</div></pre></td></tr></table></figure>]]></content>
    
    <summary type="html">
    
      &lt;blockquote&gt;
&lt;p&gt;快速排序与背包问题相关算法分析&lt;/p&gt;
&lt;/blockquote&gt;
&lt;hr&gt;
    
    </summary>
    
    
      <category term="Python" scheme="http://feiyang.li/tags/Python/"/>
    
      <category term="Algorithms" scheme="http://feiyang.li/tags/Algorithms/"/>
    
  </entry>
  
  <entry>
    <title>Coursera ML(13)-SVM(Support Vector Machines)</title>
    <link href="http://feiyang.li/2017/05/22/Coursera-ML-13/index.html"/>
    <id>http://feiyang.li/2017/05/22/Coursera-ML-13/index.html</id>
    <published>2017-05-22T12:25:02.000Z</published>
    <updated>2017-05-22T12:32:31.116Z</updated>
    
    <content type="html"><![CDATA[<script src="/assets/js/DPlayer.min.js"> </script><script src="/assets/js/APlayer.min.js"> </script><blockquote>
<p>本节笔记对应<a href="https://www.coursera.org/learn/machine-learning/home/week/7" target="_blank" rel="external">第七周课程</a>和西瓜书<strong>第六章 支持向量机</strong>，主要讲解了支持向量机。</p>
</blockquote>
<hr>
<a id="more"></a>
<h1 id="SVM介绍"><a href="#SVM介绍" class="headerlink" title="SVM介绍"></a>SVM介绍</h1><p>这一部分Coursera上的内容我<strong>并没有听懂</strong>，一头雾水。</p>
<h2 id="数学上的定义"><a href="#数学上的定义" class="headerlink" title="数学上的定义"></a>数学上的定义</h2><h3 id="逻辑回归模型"><a href="#逻辑回归模型" class="headerlink" title="逻辑回归模型"></a>逻辑回归模型</h3><p>从最简单的逻辑回国模型说起<br> $$\begin{align*}& h_\theta (x) = g ( \theta^T x ) \newline \newline& z = \theta^T x \newline& g(z) = \dfrac{1}{1 + e^{-z}}\end{align*}$$<br>简单点，也就是<br>$$h_\theta (x) = \dfrac{1}{1 + e^{-\theta^T x}} $$<br>其图像为：<br><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170522/142916098.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure><br>我们期望的的是底下那两行，而实际给我们的这个sigmoid函数，好像并没有那么完美，这是我们进行少许的处理。</p>
<h3 id="将模型进行抽象"><a href="#将模型进行抽象" class="headerlink" title="将模型进行抽象"></a>将模型进行抽象</h3><p>原本的 $h_\theta (x) = log  \dfrac{1}{1 + e^{-\theta^T x}}$ 或者 $h_\theta (x) = log(1-\dfrac{1}{1 + e^{-\theta^T x}})$ 的图像应该是底下的黑线。<br><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170522/143132931.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure></p>
<p>这里，我们定义一个新的新的符号<strong>cost</strong>，就是上边那个紫色的两条直线组成的东西。<br><strong>为什幺这么做</strong>，食屎啦，我怎么知道为什么这么做，先留着这个问题。</p>
<p>改造一番之后，我们得到了<br><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170522/145216127.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure></p>
<p>这个就是支持向量机的数学定义，去他妈的，什么鬼东西，干什么了嘛，SVM就被定义出来了。<br><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170522/121959778.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure></p>
<h2 id="直观化的定义"><a href="#直观化的定义" class="headerlink" title="直观化的定义"></a>直观化的定义</h2><p>给定这样一个样本集$D={X_1,y_1},{X_2,y_2},{X_3,y_3}……{X_m,y_m},y_i \in(-1,+1)$<br><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170522/150009680.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure><br>我们可以找到很多条直线分割这两个区域<br><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170522/151700413.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="mark" title="">
                </div>
                <div class="image-caption">mark</div>
            </figure><br>那么我们应该努力去找到哪一个呢？直观上看，那个正中间的黑线看起来很不错。</p>
<p><strong>因为该划分超平面对训练样本的局部扰动容忍度最好。</strong></p>
<ul>
<li>划分超平面可以定义为如下线性方程组：<br>$$w^Tx+b=0$$<br>其中$w=(w_1;w_2…;w_d;)$为法向量，决定平面的方向，$b$是位移量。<br>将其标记为：$(w,b)$，样本中任一点到超平面$(w,b)$的距离可以表示为：<br>$$r=    \dfrac{|w^T+b|}{||w||}$$</li>
<li>假定成功分类，那： $$ \left\{
\begin{aligned}
w^T+b \geq1,   y_i=+1\\
w^T+b \leq-1,   y_i=-1
\end{aligned}
\right.
$$ </li>
<li><p>两个异类支持向量到超平面的距离之和(间隔)为：<br>$$\gamma= \dfrac 2{||w||}$$<br>那么如果我们可以找到最大的间隔，是不是就可以使这个分割线处于最中间的位置。</p>
</li>
<li><p>使距离最小化<br>我们换一个表达方式</p>
</li>
</ul>
<p>$$\begin{aligned}<br>\min \limits_{w,b} \dfrac   12||w||^2 \<br>s.t. y_i(w^Tx+b)  \geq 1<br>\end{aligned}<br>$$<br>这就是支持向量机的基本型</p>
<p>#万能的拉格朗日－对偶问题</p>
<p>又有周boss的“简单”的数学推导了：这一次是和拉格朗日一起。基本SVM的优化目标本身是一个凸二次规划问题，可以利用优化计算包求解，但是据说拉格朗日有更高效的解法：<br>对上一节最后式子使用拉格朗日乘子法就可以得到它的对偶问题 （dual problem），对偶问题出现在线性规划中，每一个求极小的线性规划问题都有一个求极大的线性规划问题互称对偶问题，解决了一个就对应解决了其对偶问题。具体到这里，对式子的每条约束添加非负拉格朗日乘子$\alpha_i$，拉格朗日函数是<br> $$L(\boldsymbol{w},b,\boldsymbol{\alpha})=\frac{1}{2}\|\boldsymbol{w}\|^2+\sum_{i=1}^m{\alpha_i(1-y_i(\boldsymbol{w}^T\boldsymbol{x}_i+b))}$$<br>对其去$\boldsymbol{w}$和$b$的偏导为0可得到对偶问题：<br> $$max_\boldsymbol{\alpha}\ \sum_{i=1}^m\alpha_i-\frac{1}{2}\sum_{i=1}^m\sum_{j=1}^m\alpha_i\alpha_jy_iy_j\boldsymbol{x}_i^T\boldsymbol{x}_j$$$$s.t.\ \sum_{i=1}^m{\alpha_iy_i}=0,\ \alpha_i\geq0,\ i=1,2,…,m$$<br>这又是一个二次规划问题，但通用解法规模正比于样本数，在大规模数据上开销很大，可以考虑其它高效算法如SMO (Sequential Minimal Optimization)。SMO的简单描述是其不断重复以下两个步骤直至收敛：</p>
<p>选取一对需更新的变量$\alpha_i$和$\alpha_j$<br>固定$\alpha_i$和$\alpha_j$以外的参数，求解对偶问题更新后的$\alpha_i$和$\alpha_j$。</p>
<p>SMO算法在一时间也不能得到完整的理解，希望能作为学习二次规划问题的一个复杂例子在之后研究。</p>
<h1 id="核函数－核函数"><a href="#核函数－核函数" class="headerlink" title="核函数－核函数"></a>核函数－核函数</h1><h2 id="为什么要使用核函数"><a href="#为什么要使用核函数" class="headerlink" title="为什么要使用核函数"></a>为什么要使用核函数</h2><p>原始样本空间也许并不存在一个可以正确划分两类样本的超平面，一个比较好的解决办法，试讲当前原始空间映射到一个高维空间去。根据<strong>某些大佬的研究</strong>，如果原始样本空间是有限的，那么总可以找到一个高维特性空间使样本可分。</p>
<p>其实这个问题在西瓜书上说的非常清楚，但是笔记却无法做，主要是相关的图我找不出来，好气哦。</p>
<h2 id="详细介绍"><a href="#详细介绍" class="headerlink" title="详细介绍"></a>详细介绍</h2><p>取映射函数$\phi$，特征空间划分超平面模型表示为：<br>$$f(x)=\boldsymbol{w}^T\phi(x)+b$$<br>类似的有问题变为了：<br> $$min_{\boldsymbol{w},b} \frac{1}{2}\|\boldsymbol{w}\|^2$$$$s.t.\ y_i(\boldsymbol{w}^T\phi(\boldsymbol{x}_i)+b)\geq1,\ i=1,2,…,m$$<br>其对偶问题是：<br> $$max_\boldsymbol{\alpha}\ \sum_{i=1}^m\alpha_i-\frac{1}{2}\sum_{i=1}^m\sum_{j=1}^m\alpha_i\alpha_jy_iy_j\phi(\boldsymbol{x}_i^T)\phi(\boldsymbol{x}_j)$$$$s.t.\ \sum_{i=1}^m{\alpha_iy_i}=0,\ \alpha_i\geq0,\ i=1,2,…,m$$ </p>
<p>这中间涉及了$\phi(\boldsymbol{x}_i^T)\phi(\boldsymbol{x}_j)$，是样本映射到高维空间后的内积，由于维数可能很高甚至无穷，其计算很困难。此时我们就需要找一个函数，使得：<br>$$\kappa(\boldsymbol{x_i},\boldsymbol{x_j})=\phi(\boldsymbol{x}_i^T)\phi(\boldsymbol{x}_j)$$<br>称作核函数 (kernal function)。<br>有定理表明：只要一个对称函数对应的核矩阵半正定，它就能作为核函数使用。几种常用的核函数有：</p>
<table>
<thead>
<tr>
<th>名称</th>
<th style="text-align:center">表达式</th>
<th style="text-align:left">参数</th>
</tr>
</thead>
<tbody>
<tr>
<td>线性核</td>
<td style="text-align:center">$\kappa(\boldsymbol{x}_i,\boldsymbol{x}_j)=\boldsymbol{x}_i^T\boldsymbol{x}_j$</td>
<td style="text-align:left"></td>
</tr>
<tr>
<td>多项式核</td>
<td style="text-align:center">$\kappa(\boldsymbol{x}_i,\boldsymbol{x}_j)=(\boldsymbol{x}_i^T\boldsymbol{x}_j)^d$</td>
<td style="text-align:left">$d \geq 1$为多项式的次数</td>
</tr>
<tr>
<td>高斯核</td>
<td style="text-align:center">$\kappa(\boldsymbol{x}_i,\boldsymbol{x}_j)=\exp(-\frac{∥\boldsymbol{x}_i-\boldsymbol{x}_j∥^2}{2\sigma^2})$</td>
<td style="text-align:left">$\sigma \geq 0$为高斯核的带宽</td>
</tr>
<tr>
<td>拉普拉斯核</td>
<td style="text-align:center">$\kappa(\boldsymbol{x}_i,\boldsymbol{x}_j)=\exp(-\frac{∥\boldsymbol{x}_i-\boldsymbol{x}_j∥}{\sigma})$</td>
<td style="text-align:left">$\sigma \geq 0$</td>
</tr>
<tr>
<td>Sigmoid核</td>
<td style="text-align:center">$\kappa(\boldsymbol{x}_i,\boldsymbol{x}_j)=\tanh(\beta\boldsymbol{x}_i^T\boldsymbol{x}_j+\theta)$</td>
<td style="text-align:left">$\tanh$为双曲正切函数，$\beta&gt;0,\theta&lt;0$</td>
</tr>
</tbody>
</table>
<p>多个核函数的线性组合，两个核函数的直积均为核函数，以及若$\kappa_1$为核函数，对于任意函数$g(\boldsymbol{x})$有：$\kappa(\boldsymbol{x},\boldsymbol{z})=g(\boldsymbol{x})\kappa_1(\boldsymbol{x},\boldsymbol{z})g(\boldsymbol{z})$也是核函数<br>如何选择核函数将样本映射到真正显示其分布规律的高维空间非常重要。</p>
<h1 id="软间隔与正则化"><a href="#软间隔与正则化" class="headerlink" title="软间隔与正则化"></a>软间隔与正则化</h1><p>实际上很多问题并不能简单地归结为线性可分线性不可分，一些问题的样本空间看来是线性可分的，但总有几个样本跑到了敌对阵营，就为此强行升高维度显然得不偿失。于是引入了软间隔 (soft margin)：允许有若干个样本被线性空间划分错误。此时的优化目标变为：<br> $$min_{\boldsymbol{w},b}\ \frac{1}{2}\|\boldsymbol{w}\|^2+C\sum_{i=1}^m\ell_{0/1}(y_i(\boldsymbol{w}^T\boldsymbol{x}_i+b)-1)$$<br>其中$C$是一个常数，可以理解问允许多少个样本被分错，而$\ell_{0/1}$是0/1损失函数，当样本被分错时，它的取值为1，否则为0。这个函数直观，简单，然而非凸，非连续，常用一些替代函数：</p>
<p>hinge损失：$\ell<em>{hinge}(z)=\max(0,1,1-z)$<br>指数损失 (exponential loss)：$\ell</em>{exp}(z)=\exp(-z)$<br>对率损失 (logistic loss)：$\ell_{log}(z)=\log(1+\exp(-z))$</p>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170522/200058441.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="三种常见的替代损失函数对应图像为" title="">
                </div>
                <div class="image-caption">三种常见的替代损失函数对应图像为</div>
            </figure>
<ul>
<li>红色为：hinge损失</li>
<li>蓝色为：指数损失</li>
<li>黑色为：对率损失<br>这个地方跟最开始的时候提到的图像好像，只不过Andrew直接就把结果说出来了，并没有解释为什么罢了。</li>
</ul>
<p>代入优化目标后同样可以找到它们的对偶问题进行求解。实际上此时的优化目标可以看作两部分：第一项描述划分超平面的间隔大小，第二项描述数据集上的误差。对于误差还有更一般的形式： $$min_{f}\ \Omega(f)+C\sum_{i=1}^m\ell(f(\boldsymbol{x}_i),y_i)$$ 其中$\Omega(f)$称为结构风险，后一项称为经验风险，$C$用于对二者进行折中。结构风险方便引入领域知识和用户意图：用户希望得到何种性质的模型，同时它也有助于削减假设空间。这个式子称为正则化 (regularization)。</p>
<h1 id="回归问题－支持向量回归"><a href="#回归问题－支持向量回归" class="headerlink" title="回归问题－支持向量回归"></a>回归问题－支持向量回归</h1><p>普通的回归问题计算损失为函数预测值与真实值的差，而使用支持向量回归 (Support Vector Regression, SVR)，我们容忍$\epsilon$的误差，那就相当于在预测函数两端建立了一个宽为$2\epsilon$的隔离带，在此间隔带种的样本被默认为预测正确，而在此间隔带外的样本计算它的真实值与预测函数得到值的差作为损失，就此SVR问题可形式化为：<br> $$min_{\boldsymbol{w},b}\ \frac{1}{2}\|\boldsymbol{w}\|^2+C\sum_{i=1}^m\ell_{\epsilon}(f(\boldsymbol{x}_i)-y_i)$$ </p>
<p>其中$C&gt;0$是一个常数,$l_{0/1}$是”0/1”损失函数<br> $$\begin{equation}l_{0/1}=\left\{\begin{array}{ll}
1,\,\text{if}\quad z<0\\ 0,otherwise="" \end{array}\right.\end{equation}$$="" <="" p="">
<p>常见的损失函数有hinge损失,指数损失,对率损失(见P130).</p>
<p>采用hinge损失,引入”松弛变量”(slack variables),上式可重写为<br> $$\begin{equation}\begin{split}
\min_{\mathbf w,b,\xi}&\frac{1}{2}\|\mathbf w\|^2+C\sum_{i=1}^m\xi_i\\
\text{s.t.}&\,y_i(\mathbf{w^Tx_i}+b)\ge1-\xi_i\\
&\xi_i\ge0,\,i=1,2,\cdots,m.
\end{split}\end{equation}$$ </p>
<p>对偶问题推导以及解稀疏性证明见P132.</p>
<p>支持向量引申到回归问题上即可得到支持向量回归(Support Vector Regression, SVR).SVR假设我们能容忍$f(\mathbf x)$与$y$之间最多有$\epsilon$的偏差,即仅当$f(\mathbf x)$与$y$之间的差别绝对值大于$\epsilon$时才计算损失.SVR问题可化为<br> $$\begin{equation}\begin{split}
\min_{\mathbf w,b}\frac{1}{2}\|\mathbf w\|^2+C\sum_{i=1}^ml_\epsilon(f(\mathbf x_i)-y_i)
\end{split}\end{equation}$$ </p>
<p>其中 $ C $为正则化常数,$l_\epsilon$为$\epsilon$-不敏感损失函数</p>
 $$\begin{equation}l_\epsilon=\left\{\begin{array}{ll}
0,\quad\text{if}|z|\le\epsilon\\
|z|-\epsilon,otherwise
\end{array}\right.\end{equation}$$ 
<p>对偶问题推导以及KKT条件证明见P135-137.</p>
<h1 id="核-nuclear-方法"><a href="#核-nuclear-方法" class="headerlink" title="核(nuclear)方法"></a>核(nuclear)方法</h1><p>给定训练样本,不考虑偏移项$b$,则无论SVM还是SVR,学得的模型总能表示成核函数$\kappa(\mathbf{x_i,x_j})$的线性组合.因此有更一般的”表示定理”(representer theorem).P137。</p>
<p>由此人们发展了一系列基于核函数的学习方法，统称为核方法 (kernel methods)。最常见的就是引入核函数来将线性学习器拓展为非线性学习器，从而得到核线性判别分析 (Kernelized Linear Discriminant Analysis)。</p>
</0\\></p>]]></content>
    
    <summary type="html">
    
      &lt;blockquote&gt;
&lt;p&gt;本节笔记对应&lt;a href=&quot;https://www.coursera.org/learn/machine-learning/home/week/7&quot;&gt;第七周课程&lt;/a&gt;和西瓜书&lt;strong&gt;第六章 支持向量机&lt;/strong&gt;，主要讲解了支持向量机。&lt;/p&gt;
&lt;/blockquote&gt;
&lt;hr&gt;
    
    </summary>
    
      <category term="Coursera ML" scheme="http://feiyang.li/categories/Coursera-ML/"/>
    
    
      <category term="Machine Learning" scheme="http://feiyang.li/tags/Machine-Learning/"/>
    
      <category term="读书笔记" scheme="http://feiyang.li/tags/%E8%AF%BB%E4%B9%A6%E7%AC%94%E8%AE%B0/"/>
    
  </entry>
  
  <entry>
    <title>Django 笔记之三 Django 模板</title>
    <link href="http://feiyang.li/2017/05/21/django3/index.html"/>
    <id>http://feiyang.li/2017/05/21/django3/index.html</id>
    <published>2017-05-21T08:31:52.000Z</published>
    <updated>2017-05-21T08:43:14.437Z</updated>
    
    <content type="html"><![CDATA[<script src="/assets/js/DPlayer.min.js"> </script><script src="/assets/js/APlayer.min.js"> </script><blockquote>
<p>在前面的几节中我们都是用简单的 django.http.HttpResponse 来把内容显示到网页上，本节将讲解如何使用渲染模板的方法来显示内容。</p>
</blockquote>
<hr>
<a id="more"></a>
<h1 id="模板"><a href="#模板" class="headerlink" title="模板"></a>模板</h1><p>网站模板的设计，一般的，我们做网站有一些通用的部分，比如 导航，底部，访问统计代码等等.<br>比如<strong>nav.html, bottom.html, tongji.html</strong></p>
<p>可以写一个 base.html 来包含这些通用文件（include)<br><figure class="highlight html"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div><div class="line">10</div><div class="line">11</div><div class="line">12</div><div class="line">13</div><div class="line">14</div><div class="line">15</div><div class="line">16</div><div class="line">17</div><div class="line">18</div><div class="line">19</div></pre></td><td class="code"><pre><div class="line"><span class="meta">&lt;!DOCTYPE html&gt;</span></div><div class="line"><span class="tag">&lt;<span class="name">html</span>&gt;</span></div><div class="line"><span class="tag">&lt;<span class="name">head</span>&gt;</span></div><div class="line">    <span class="tag">&lt;<span class="name">title</span>&gt;</span>&#123;% block title %&#125;默认标题&#123;% endblock %&#125; - 自强学堂<span class="tag">&lt;/<span class="name">title</span>&gt;</span></div><div class="line"><span class="tag">&lt;/<span class="name">head</span>&gt;</span></div><div class="line"><span class="tag">&lt;<span class="name">body</span>&gt;</span></div><div class="line"> </div><div class="line">&#123;% include 'nav.html' %&#125;</div><div class="line"> </div><div class="line">&#123;% block content %&#125;</div><div class="line"><span class="tag">&lt;<span class="name">div</span>&gt;</span>这里是默认内容，所有继承自这个模板的，如果不覆盖就显示这里的默认内容。<span class="tag">&lt;/<span class="name">div</span>&gt;</span></div><div class="line">&#123;% endblock %&#125;</div><div class="line"> </div><div class="line">&#123;% include 'bottom.html' %&#125;</div><div class="line"> </div><div class="line">&#123;% include 'tongji.html' %&#125;</div><div class="line"> </div><div class="line"><span class="tag">&lt;/<span class="name">body</span>&gt;</span></div><div class="line"><span class="tag">&lt;/<span class="name">html</span>&gt;</span></div></pre></td></tr></table></figure></p>
<p>如果需要，写足够多的 <strong>block</strong> 以便继承的模板可以重写该部分，<strong>include</strong> 是包含其它文件的内容，就是把一些网页共用的部分拿出来，重复利用，改动的时候也方便一些，还可以把广告代码放在一个单独的<strong>html</strong>中，改动也方便一些，在用到的地方<strong>include</strong>进去。其它的页面继承自<strong> base.html </strong>就好了，继承后的模板也可以在<strong> block</strong> 块中<strong> include </strong>其它的模板文件。</p>
<p>比如我们的首页 <strong>home.html</strong>，继承或者说扩展(extends)原来的 <strong>base.html</strong>，可以简单这样写，重写部分代码（默认值的那一部分不用改）<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div></pre></td><td class="code"><pre><div class="line">&#123;% extends &apos;base.html&apos; %&#125;</div><div class="line"> </div><div class="line">&#123;% block title %&#125;欢迎光临首页&#123;% endblock %&#125;</div><div class="line"> </div><div class="line">&#123;% block content %&#125;</div><div class="line">&#123;% include &apos;ad.html&apos; %&#125;</div><div class="line">这里是首页，欢迎光临</div><div class="line">&#123;% endblock %&#125;</div></pre></td></tr></table></figure></p>
<p>注意：模板一般放在app下的templates中，Django会自动去这个文件夹中找。但 假如我们每个app的templates中都有一个 index.html，当我们在views.py中使用的时候，直接写一个<strong> render(request, ‘index.html’)</strong>，Django 能不能找到当前 app 的 templates 文件夹中的 index.html 文件夹呢?（<strong>答案是不一定能，有可能找错</strong>）</p>
<p>##Django 模板查找机制<br> Django 查找模板的过程是在每个 app 的 templates 文件夹中找（而不只是当前 app 中的代码只在当前的 app 的 templates 文件夹中找）。各个 app 的 templates 形成一个文件夹列表，Django 遍历这个列表，一个个文件夹进行查找，当在某一个文件夹找到的时候就停止，所有的都遍历完了还找不到指定的模板的时候就是 Template Not Found （过程类似于Python找包）。这样设计有利当然也有弊，有利是的地方是一个app可以用另一个app的模板文件，弊是有可能会找错了。所以我们使用的时候在 templates 中建立一个 app 同名的文件夹，这样就好了。</p>
<p>这样，使用的时候，模板就是 “blog/index.html” 和 “mmmwhy/index.html” 这样有app作为名称的一部分，就不会混淆。</p>
<h1 id="Django-模板进阶"><a href="#Django-模板进阶" class="headerlink" title="Django 模板进阶"></a>Django 模板进阶</h1><p>本节主要讲 Django模板中的循环，条件判断，常用的标签，过滤器的使用。</p>
<h2 id="显示一个基本的字符串在网页上"><a href="#显示一个基本的字符串在网页上" class="headerlink" title="显示一个基本的字符串在网页上"></a>显示一个基本的字符串在网页上</h2><ul>
<li>views.py<figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div></pre></td><td class="code"><pre><div class="line"><span class="function"><span class="keyword">def</span> <span class="title">home</span><span class="params">(request)</span>:</span></div><div class="line">    string = <span class="string">u"我是一个小胖子"</span></div><div class="line">    <span class="keyword">return</span> render(request, <span class="string">'home.html'</span>, &#123;<span class="string">'string'</span>: string&#125;)</div></pre></td></tr></table></figure>
</li>
</ul>
<p>在视图中我们传递了一个字符串名称是 string 到模板 home.html，在模板中这样使用它：</p>
<ul>
<li>home.html<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">&#123;&#123; string &#125;&#125;</div></pre></td></tr></table></figure>
</li>
</ul>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170521/113457331.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<h2 id="基本的-for-循环-和-List内容的显示"><a href="#基本的-for-循环-和-List内容的显示" class="headerlink" title="基本的 for 循环 和 List内容的显示"></a>基本的 for 循环 和 List内容的显示</h2><ul>
<li><p>views.py</p>
<figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div></pre></td><td class="code"><pre><div class="line"><span class="function"><span class="keyword">def</span> <span class="title">home</span><span class="params">(request)</span>:</span></div><div class="line">    TutorialList = [<span class="string">"HTML"</span>, <span class="string">"CSS"</span>, <span class="string">"jQuery"</span>, <span class="string">"Python"</span>, <span class="string">"Django"</span>]</div><div class="line">    <span class="keyword">return</span> render(request, <span class="string">'home.html'</span>, &#123;<span class="string">'TutorialList'</span>: TutorialList&#125;)</div></pre></td></tr></table></figure>
</li>
<li><p>home.html</p>
<figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div></pre></td><td class="code"><pre><div class="line">&#123;% <span class="keyword">for</span> i <span class="keyword">in</span> TutorialList %&#125;</div><div class="line">&#123;&#123; i &#125;&#125;</div><div class="line">&#123;% endfor %&#125; <span class="comment">#for 循环要有一个结束标记</span></div></pre></td></tr></table></figure>
</li>
</ul>
<h2 id="显示字典中内容："><a href="#显示字典中内容：" class="headerlink" title="显示字典中内容："></a>显示字典中内容：</h2><ul>
<li><p>views.py</p>
<figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div></pre></td><td class="code"><pre><div class="line"><span class="function"><span class="keyword">def</span> <span class="title">home</span><span class="params">(request)</span>:</span></div><div class="line">    info_dict = &#123;<span class="string">'site'</span>: <span class="string">u'肥羊'</span>, <span class="string">'content'</span>: <span class="string">u'小姿势'</span>&#125;</div><div class="line">    <span class="keyword">return</span> render(request, <span class="string">'home.html'</span>, &#123;<span class="string">'info_dict'</span>: info_dict&#125;)</div></pre></td></tr></table></figure>
</li>
<li><p>home.html</p>
<figure class="highlight html"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">站点：&#123;&#123; info_dict.site &#125;&#125; 内容：&#123;&#123; info_dict.content &#125;&#125;</div></pre></td></tr></table></figure>
</li>
</ul>
<p>或者可以遍历字典内容：<br><figure class="highlight html"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div></pre></td><td class="code"><pre><div class="line">&#123;% for key, value in info_dict.items %&#125;</div><div class="line">    &#123;&#123; key &#125;&#125;: &#123;&#123; value &#125;&#125;</div><div class="line">&#123;% endfor %&#125;</div></pre></td></tr></table></figure></p>
<h2 id="在模板进行-条件判断和-for-循环的详细操作："><a href="#在模板进行-条件判断和-for-循环的详细操作：" class="headerlink" title="在模板进行 条件判断和 for 循环的详细操作："></a>在模板进行 条件判断和 for 循环的详细操作：</h2><ul>
<li><p>views.py</p>
<figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div></pre></td><td class="code"><pre><div class="line"><span class="function"><span class="keyword">def</span> <span class="title">home</span><span class="params">(request)</span>:</span></div><div class="line">    List = map(str, range(<span class="number">100</span>))<span class="comment"># 一个长度为100的 List</span></div><div class="line">    <span class="keyword">return</span> render(request, <span class="string">'home.html'</span>, &#123;<span class="string">'List'</span>: List&#125;)</div></pre></td></tr></table></figure>
</li>
<li><p>home.html</p>
<figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div></pre></td><td class="code"><pre><div class="line">&#123;% <span class="keyword">for</span> item <span class="keyword">in</span> List %&#125;</div><div class="line">    &#123;&#123; item &#125;&#125;&#123;% <span class="keyword">if</span> <span class="keyword">not</span> forloop.last %&#125;,&#123;% endif %&#125;</div><div class="line">&#123;% endfor %&#125;</div></pre></td></tr></table></figure>
</li>
</ul>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170521/134312852.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<h2 id="模板中-获取当前网址，当前用户等"><a href="#模板中-获取当前网址，当前用户等" class="headerlink" title="模板中 获取当前网址，当前用户等"></a>模板中 获取当前网址，当前用户等</h2><ul>
<li><p>获取当前用户</p>
<figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div></pre></td><td class="code"><pre><div class="line">&#123;% <span class="keyword">if</span> request.user.is_authenticated %&#125;</div><div class="line">    &#123;&#123; request.user.username &#125;&#125;，您好！</div><div class="line">&#123;% <span class="keyword">else</span> %&#125;</div><div class="line">    请登陆，这里放登陆链接</div><div class="line">&#123;% endif %&#125;</div></pre></td></tr></table></figure>
</li>
<li><p>获取当前网址</p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">&#123;&#123; request.path &#125;&#125;</div></pre></td></tr></table></figure>
</li>
<li><p>获取当前 GET 参数：</p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">&#123;&#123; request.GET.urlencode &#125;&#125;</div></pre></td></tr></table></figure></li>
</ul>
]]></content>
    
    <summary type="html">
    
      &lt;blockquote&gt;
&lt;p&gt;在前面的几节中我们都是用简单的 django.http.HttpResponse 来把内容显示到网页上，本节将讲解如何使用渲染模板的方法来显示内容。&lt;/p&gt;
&lt;/blockquote&gt;
&lt;hr&gt;
    
    </summary>
    
      <category term="Django" scheme="http://feiyang.li/categories/Django/"/>
    
    
      <category term="教程" scheme="http://feiyang.li/tags/%E6%95%99%E7%A8%8B/"/>
    
      <category term="Django" scheme="http://feiyang.li/tags/Django/"/>
    
  </entry>
  
  <entry>
    <title>Django 笔记之二 视图与网址</title>
    <link href="http://feiyang.li/2017/05/20/django2/index.html"/>
    <id>http://feiyang.li/2017/05/20/django2/index.html</id>
    <published>2017-05-20T08:06:53.000Z</published>
    <updated>2017-05-21T08:37:15.595Z</updated>
    
    <content type="html"><![CDATA[<script src="/assets/js/DPlayer.min.js"> </script><script src="/assets/js/APlayer.min.js"> </script><blockquote>
<p><a href="http://feiyang.li/2017/05/20/django1/index.html">上节</a>已经提到新建app，这节稍微深入一些</p>
</blockquote>
<hr>
<a id="more"></a>
<h1 id="新建app"><a href="#新建app" class="headerlink" title="新建app"></a>新建app</h1><h2 id="基本步骤"><a href="#基本步骤" class="headerlink" title="基本步骤"></a>基本步骤</h2><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">python manage.py startapp learn # learn 是一个app的名称</div></pre></td></tr></table></figure>
<p>之后大概是这个样子，也许你们会发现我用的其实是<strong>世界上最好的python IDE</strong>~<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div><div class="line">10</div><div class="line">11</div><div class="line">12</div><div class="line">13</div><div class="line">14</div><div class="line">15</div><div class="line">16</div><div class="line">17</div><div class="line">18</div><div class="line">19</div><div class="line">20</div><div class="line">21</div><div class="line">22</div><div class="line">23</div><div class="line">24</div><div class="line">25</div><div class="line">26</div><div class="line">27</div><div class="line">28</div><div class="line">29</div><div class="line">30</div><div class="line">31</div><div class="line">32</div><div class="line">33</div><div class="line">34</div><div class="line">35</div><div class="line">36</div><div class="line">37</div><div class="line">38</div><div class="line">39</div><div class="line">40</div><div class="line">41</div><div class="line">42</div><div class="line">43</div><div class="line">44</div><div class="line">45</div><div class="line">46</div><div class="line">47</div><div class="line">48</div><div class="line">49</div><div class="line">50</div><div class="line">51</div><div class="line">52</div><div class="line">53</div><div class="line">54</div></pre></td><td class="code"><pre><div class="line">E:.</div><div class="line">│  db.sqlite3</div><div class="line">│  manage.py</div><div class="line">│  Python</div><div class="line">│</div><div class="line">├─.idea</div><div class="line">│  │  blog.iml</div><div class="line">│  │  misc.xml</div><div class="line">│  │  modules.xml</div><div class="line">│  │  workspace.xml</div><div class="line">│  │</div><div class="line">│  └─inspectionProfiles</div><div class="line">│          profiles_settings.xml</div><div class="line">│</div><div class="line">├─blog</div><div class="line">│  │  settings.py</div><div class="line">│  │  settings.pyc</div><div class="line">│  │  urls.py</div><div class="line">│  │  urls.pyc</div><div class="line">│  │  wsgi.py</div><div class="line">│  │  wsgi.pyc</div><div class="line">│  │  __init__.py</div><div class="line">│  │  __init__.pyc</div><div class="line">│  │</div><div class="line">│  └─__pycache__</div><div class="line">│          settings.cpython-35.pyc</div><div class="line">│          urls.cpython-35.pyc</div><div class="line">│          wsgi.cpython-35.pyc</div><div class="line">│          __init__.cpython-35.pyc</div><div class="line">│</div><div class="line">└─my_blog</div><div class="line">    │  admin.py</div><div class="line">    │  admin.pyc</div><div class="line">    │  apps.py</div><div class="line">    │  models.py</div><div class="line">    │  models.pyc</div><div class="line">    │  tests.py</div><div class="line">    │  views.py</div><div class="line">    │  views.pyc</div><div class="line">    │  __init__.py</div><div class="line">    │  __init__.pyc</div><div class="line">    │</div><div class="line">    ├─migrations</div><div class="line">    │  │  __init__.py</div><div class="line">    │  │  __init__.pyc</div><div class="line">    │  │</div><div class="line">    │  └─__pycache__</div><div class="line">    │          __init__.cpython-35.pyc</div><div class="line">    │</div><div class="line">    └─__pycache__</div><div class="line">            admin.cpython-35.pyc</div><div class="line">            models.cpython-35.pyc</div><div class="line">            views.cpython-35.pyc</div><div class="line">            __init__.cpython-35.pyc</div></pre></td></tr></table></figure></p>
<h2 id="声明my-blog"><a href="#声明my-blog" class="headerlink" title="声明my_blog"></a>声明<strong>my_blog</strong></h2><p>我们需要在<strong>settings.py</strong>中去声明一下<strong>my_blog</strong></p>
<figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div></pre></td><td class="code"><pre><div class="line">INSTALLED_APPS = [</div><div class="line">    <span class="string">'django.contrib.admin'</span>,</div><div class="line">    <span class="string">'django.contrib.auth'</span>,</div><div class="line">    <span class="string">'django.contrib.contenttypes'</span>,</div><div class="line">    <span class="string">'django.contrib.sessions'</span>,</div><div class="line">    <span class="string">'django.contrib.messages'</span>,</div><div class="line">    <span class="string">'django.contrib.staticfiles'</span>,</div><div class="line">    <span class="string">'my_blog'</span>,</div><div class="line">]</div></pre></td></tr></table></figure>
<p>新建的 app 如果不加到 INSTALL_APPS 中的话, django 就不能自动找到app中的模板文件(app-name/templates/下的文件)和静态文件(app-name/static/中的文件) 。</p>
<h2 id="定义视图内容"><a href="#定义视图内容" class="headerlink" title="定义视图内容"></a>定义视图内容</h2><p>将views.py修改为我们希望展示的内容：<br><figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div></pre></td><td class="code"><pre><div class="line"><span class="comment">#coding: utf-8</span></div><div class="line"><span class="keyword">from</span> django.http <span class="keyword">import</span> HttpResponse</div><div class="line"><span class="function"><span class="keyword">def</span> <span class="title">index</span><span class="params">(request)</span>:</span></div><div class="line">    <span class="keyword">return</span> HttpResponse(<span class="string">u'你好哇，李肥羊，希望你可以瘦下来！'</span>)</div></pre></td></tr></table></figure></p>
<p>使用<code>python manage.py runserver</code>，启动django服务</p>
<ul>
<li><p>第一行是声明编码为utf-8, 因为我们在代码中用到了中文,如果不声明就报错.</p>
</li>
<li><p>第二行引入HttpResponse，它是用来向网页返回内容的，就像Python中的 print 一样，只不过 HttpResponse 是把内容显示到网页上。</p>
</li>
</ul>
<p>我们定义了一个index()函数，第一个参数必须是 request，与网页发来的请求有关，request 变量里面包含get或post的内容，用户浏览器，系统等信息在里面（后面会讲，先了解一下就可以）。</p>
<p>函数返回了一个 HttpResponse 对象，可以经过一些处理，最终显示几个字到网页上。<br><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170520/103039383.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure></p>
<p>如果你前端技术非常好的话，可以直接把整个页面return回去。。。</p>
<p>那问题来了，我们访问什么网址才能看到刚才写的这个函数呢？怎么让网址和函数关联起来呢？</p>
<h2 id="定义视图函数相关的URL-网址"><a href="#定义视图函数相关的URL-网址" class="headerlink" title="定义视图函数相关的URL(网址)"></a>定义视图函数相关的URL(网址)</h2><p>即规定 访问什么网址对应什么内容，我们打开 urls.py 这个文件, 修改其中的代码:<br><figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div></pre></td><td class="code"><pre><div class="line"><span class="keyword">from</span> django.conf.urls <span class="keyword">import</span> url</div><div class="line"><span class="keyword">from</span> django.contrib <span class="keyword">import</span> admin</div><div class="line"><span class="keyword">from</span> my_blog <span class="keyword">import</span> views <span class="keyword">as</span> my_blog_view</div><div class="line">urlpatterns = [</div><div class="line">    url(<span class="string">r'^admin/'</span>, admin.site.urls),</div><div class="line">    url(<span class="string">r'^$'</span>, my_blog_view.index),</div><div class="line">]</div></pre></td></tr></table></figure></p>
<p>先从<strong>my_blog</strong>导入<strong>views</strong>进入<strong>urls.py</strong>，然后用正则表达式，设定对应规则。如果忘了正则表达式怎么使用，也许这个<a href="http://feiyang.li/2017/02/10/Python_lambda/index.html">网页</a>会对你有帮助.</p>
<h1 id="通过链接做加减法"><a href="#通过链接做加减法" class="headerlink" title="通过链接做加减法"></a>通过链接做加减法</h1><p>跟之前的步骤一样，先修改<strong>view.py</strong>文件</p>
<h2 id="修改view-py"><a href="#修改view-py" class="headerlink" title="修改view.py"></a>修改view.py</h2><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div><div class="line">10</div><div class="line">11</div><div class="line">12</div><div class="line">13</div><div class="line">14</div></pre></td><td class="code"><pre><div class="line">def add(request):</div><div class="line">    a = request.GET[&apos;a&apos;]</div><div class="line">    b = request.GET[&apos;b&apos;]</div><div class="line">    c = int(a)+int(b)</div><div class="line">    return HttpResponse(str(c))</div><div class="line"></div><div class="line">def sub(request):</div><div class="line">    a = request.GET[&apos;a&apos;]</div><div class="line">    b = request.GET[&apos;b&apos;]</div><div class="line">    c = int(a)-int(b)</div><div class="line">    return HttpResponse(str(c))</div><div class="line">def add2(request, a, b):</div><div class="line">    c = int(a) + int(b)</div><div class="line">    return HttpResponse(str(c))</div></pre></td></tr></table></figure>
<h2 id="修改urls-py"><a href="#修改urls-py" class="headerlink" title="修改urls.py"></a>修改urls.py</h2><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div></pre></td><td class="code"><pre><div class="line">url(r&apos;^add/$&apos;, my_blog_view.add, name=&apos;add&apos;),  # 注意修改了这一行</div><div class="line">url(r&apos;^sub/$&apos;, my_blog_view.add, name=&apos;sub&apos;),  # 注意修改了这一行</div><div class="line">url(r&apos;^add/(\d+)/(\d+)/$&apos;, my_blog_view.add2, name=&apos;add2&apos;),</div></pre></td></tr></table></figure>
<h2 id="运行检查"><a href="#运行检查" class="headerlink" title="运行检查"></a>运行检查</h2><p>输入<a href="http://127.0.0.1:8000/add/?a=4&amp;b=5或者http://127.0.0.1:8000/add/?a=4&amp;b=5" target="_blank" rel="external">http://127.0.0.1:8000/add/?a=4&amp;b=5或者http://127.0.0.1:8000/add/?a=4&amp;b=5</a><br><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170520/151648247.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure><br>就可以获得答案</p>
<h1 id="URL-name返回整个页面"><a href="#URL-name返回整个页面" class="headerlink" title="URL name返回整个页面"></a>URL name返回整个页面</h1><h2 id="urls-py"><a href="#urls-py" class="headerlink" title="urls.py"></a>urls.py</h2><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">url(r&apos;^add/$&apos;, calc_views.add, name=&apos;add&apos;)</div></pre></td></tr></table></figure>
<p> 这里的 name=’add’ 是用来干什么的呢，简单说，name 可以用于在 templates, models, views ……中得到对应的网址，相当于“给网址取了个名字”，只要这个名字不变，网址变了也能通过名字获取到。</p>
<h2 id="views-py"><a href="#views-py" class="headerlink" title="views.py"></a>views.py</h2><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div></pre></td><td class="code"><pre><div class="line">from django.http import HttpResponse</div><div class="line">from django.shortcuts import render</div><div class="line"> </div><div class="line"> </div><div class="line">def index(request):</div><div class="line">    return render(request, &apos;home.html&apos;)</div></pre></td></tr></table></figure>
<h2 id="templates"><a href="#templates" class="headerlink" title="templates"></a>templates</h2><p>我们在 my_blog 这个 app 中新建一个 templates 文件夹，在templates中新建一个 home.html<br>文件 my_blog/templates/home.html 中写入以下内容（保存时用 utf8 编码）<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div><div class="line">10</div></pre></td><td class="code"><pre><div class="line">&lt;!DOCTYPE html&gt;</div><div class="line">&lt;html lang=&quot;en&quot;&gt;</div><div class="line">&lt;head&gt;</div><div class="line">    &lt;meta charset=&quot;UTF-8&quot;&gt;</div><div class="line">    &lt;title&gt;今天是个好天气&lt;/title&gt;</div><div class="line">&lt;/head&gt;</div><div class="line">&lt;body&gt;</div><div class="line">&lt;a href=&quot;/add/4/5/&quot;&gt;计算 4+5&lt;/a&gt;</div><div class="line">&lt;/body&gt;</div><div class="line">&lt;/html&gt;</div></pre></td></tr></table></figure></p>
<h2 id="运行结果"><a href="#运行结果" class="headerlink" title="运行结果"></a>运行结果</h2><p>运行开发服务器，我们访问 <a href="http://127.0.0.1:8000/" target="_blank" rel="external">http://127.0.0.1:8000/</a> 可以看到<br><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170520/153435675.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure><br>点击<strong>计算4+5</strong>，可以看到答案。</p>
<h2 id="如何用-Python-代码获取对应的网址"><a href="#如何用-Python-代码获取对应的网址" class="headerlink" title="如何用 Python 代码获取对应的网址"></a>如何用 Python 代码获取对应的网址</h2><p>···<br><a href="/add/4/5/">计算 4+5</a><br>···<br>如果这样写“死网址”，之后上线后，如果想要改变链接结构，将会很麻烦。如果有些地方没有修改过来，就不能使用了、<br>使用<code>python manage.py shell</code>，进入<code>django shell</code>调试页面.<br><figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div></pre></td><td class="code"><pre><div class="line"><span class="meta">&gt;&gt;&gt; </span><span class="keyword">from</span> django.core.urlresolvers <span class="keyword">import</span> reverse  <span class="comment"># django 1.4.x - django 1.10.x</span></div><div class="line"><span class="meta">&gt;&gt;&gt; </span><span class="keyword">from</span> django.urls <span class="keyword">import</span> reverse  <span class="comment"># django 1.10.x 新的，更加规范了</span></div><div class="line"> </div><div class="line"><span class="meta">&gt;&gt;&gt; </span>reverse(<span class="string">'add2'</span>, args=(<span class="number">4</span>,<span class="number">5</span>))</div><div class="line"><span class="string">u'/add/4/5/'</span></div><div class="line"><span class="meta">&gt;&gt;&gt; </span>reverse(<span class="string">'add2'</span>, args=(<span class="number">444</span>,<span class="number">555</span>))</div><div class="line"><span class="string">u'/add/444/555/'</span></div></pre></td></tr></table></figure></p>
<p><strong>reverse 接收 url 中的 name 作为第一个参数</strong>，我们在代码中就可以通过 reverse() 来获取对应的网址。也就是输入name返回<strong>对应网址</strong>。</p>
<p>例如：<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">&lt;a href=&quot;&#123;% url &apos;add2&apos; 4 5 %&#125;&quot;&gt; 计算 4+5&lt;/a&gt;</div></pre></td></tr></table></figure></p>
<p>上面的代码渲染成最终的页面是</p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">&lt;a href=&quot;/add/4/5/&quot;&gt;计算 4+5&lt;/a&gt;</div></pre></td></tr></table></figure>
 |这样就可以通过 {% url 'add2' 4 5 %} 获取到对应的网址 /add/4/5/ |<br><br>这样，即使我们在<strong>urls.py</strong>中修改了网址链接，<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">url(r&apos;^new_add/(\d+)/(\d+)/$&apos;, my_blog_view.add2, name=&apos;add2&apos;),</div></pre></td></tr></table></figure>
<p>注意看重点 add 变成了 new_add，但是后面的 name=’add2’ 没改，这时  |`{% url 'add2' 4 5 %}` | 就会渲染对应的网址成<code>/new_add/4/5/</code></p>
<p>用在 views.py 或 models.py 等地方的 reverse函数，同样会根据 name 对应的url获取到新的网址。</p>
<p>想要改网址的时候，修改 urls.py 中的正则表达式部分（url 参数第一部分），name 不变的前提下，其它地方都不需要修改。</p>
<h2 id="链接跳转"><a href="#链接跳转" class="headerlink" title="链接跳转"></a>链接跳转</h2><h3 id="views-py-写一个跳转的函数"><a href="#views-py-写一个跳转的函数" class="headerlink" title="views.py 写一个跳转的函数"></a>views.py 写一个跳转的函数</h3><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div></pre></td><td class="code"><pre><div class="line">from django.http import HttpResponseRedirect</div><div class="line">from django.urls import reverse  # new in django 1.10.x</div><div class="line"> </div><div class="line">def old_add2_redirect(request, a, b):</div><div class="line">    return HttpResponseRedirect(</div><div class="line">        reverse(&apos;add2&apos;, args=(a, b))</div><div class="line">    )</div></pre></td></tr></table></figure>
<h3 id="urls-py-1"><a href="#urls-py-1" class="headerlink" title="urls.py"></a>urls.py</h3><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div></pre></td><td class="code"><pre><div class="line">url(r&apos;^add/(\d+)/(\d+)/$&apos;, my_blog_view.old_add2_redirect),</div><div class="line">url(r&apos;^new_add/(\d+)/(\d+)/$&apos;, my_blog_view.add2, name=&apos;add2&apos;),</div></pre></td></tr></table></figure>
<p>即可完成跳转</p>
]]></content>
    
    <summary type="html">
    
      &lt;blockquote&gt;
&lt;p&gt;&lt;a href=&quot;http://feiyang.li/2017/05/20/django1/index.html&quot;&gt;上节&lt;/a&gt;已经提到新建app，这节稍微深入一些&lt;/p&gt;
&lt;/blockquote&gt;
&lt;hr&gt;
    
    </summary>
    
      <category term="Django" scheme="http://feiyang.li/categories/Django/"/>
    
    
      <category term="教程" scheme="http://feiyang.li/tags/%E6%95%99%E7%A8%8B/"/>
    
      <category term="Django" scheme="http://feiyang.li/tags/Django/"/>
    
  </entry>
  
  <entry>
    <title>Django 笔记之一 基本命令</title>
    <link href="http://feiyang.li/2017/05/20/django1/index.html"/>
    <id>http://feiyang.li/2017/05/20/django1/index.html</id>
    <published>2017-05-20T01:20:41.000Z</published>
    <updated>2017-05-21T08:32:12.067Z</updated>
    
    <content type="html"><![CDATA[<script src="/assets/js/DPlayer.min.js"> </script><script src="/assets/js/APlayer.min.js"> </script><p>本节主要是为了让您了解一些django最基本的命令，请尝试着记住它们，并且多多练习下，打开 Linux 或 MacOS 的 Terminal （终端）直接在 终端中输入这些命令（不是 python 的 shell中）</p>
<hr>
<a id="more"></a>
<p>如果是 windows 用 cmd（开始 搜索 cmd 或者 快捷键 win + R，输入 cmd) 直接在 cmd 上操作。</p>
<h1 id="新建一个-django-project"><a href="#新建一个-django-project" class="headerlink" title="新建一个 django project"></a>新建一个 django project</h1><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">django-admin startproject blog</div></pre></td></tr></table></figure>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://upload-images.jianshu.io/upload_images/3326314-20ef9944c7e2688d.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<p>特别是在 windows 上，如果报错，尝试用 django-admin 代替 django-admin.py 试试<br>注意 project_name 是自己的项目名称，需要为合法的 Python 包名，如不能为 1a 或 a-b。</p>
<h1 id="新建-app"><a href="#新建-app" class="headerlink" title="新建 app"></a>新建 app</h1><p>要先进入项目目录下，cd project_name 然后执行下面的命令（下同，已经在项目目录下则不需要 cd project_name）<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">python manage.py startapp my_blog</div></pre></td></tr></table></figure></p>
<p>或<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">django-admin  startapp my_blog</div></pre></td></tr></table></figure></p>
<p>效果太概是这个样子的<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div><div class="line">10</div><div class="line">11</div><div class="line">12</div><div class="line">13</div><div class="line">14</div><div class="line">15</div><div class="line">16</div><div class="line">17</div><div class="line">18</div></pre></td><td class="code"><pre><div class="line">.</div><div class="line">├── my_blog</div><div class="line">│   ├── admin.py</div><div class="line">│   ├── __init__.py</div><div class="line">│   ├── migrations</div><div class="line">│   │   └── __init__.py</div><div class="line">│   ├── models.py</div><div class="line">│   ├── tests.py</div><div class="line">│   └── views.py</div><div class="line">├── blog</div><div class="line">│   ├── __init__.py</div><div class="line">│   ├── __pycache__</div><div class="line">│   │   ├── __init__.cpython-34.pyc</div><div class="line">│   │   └── settings.cpython-34.pyc</div><div class="line">│   ├── settings.py</div><div class="line">│   ├── urls.py</div><div class="line">│   └── wsgi.py</div><div class="line">└── manage.py</div></pre></td></tr></table></figure></p>
<p>一般一个项目有多个app, 当然通用的app也可以在多个项目中使用。</p>
<p>与项目名类似 app name 也需要为合法的 Python 包名，如 blog，news, aboutus 等都是合法的 app 名称。</p>
<h1 id="同步更改数据库表或字段"><a href="#同步更改数据库表或字段" class="headerlink" title="同步更改数据库表或字段"></a>同步更改数据库表或字段</h1><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div></pre></td><td class="code"><pre><div class="line">Django 1.7.1及以上 用以下命令</div><div class="line"># 1. 创建更改的文件</div><div class="line">python manage.py makemigrations</div><div class="line"># 2. 将生成的py文件应用到数据库</div><div class="line">python manage.py migrate</div></pre></td></tr></table></figure>
<p>这种方法可以在SQL等数据库中创建与models.py代码对应的表，不需要自己手动执行SQL。</p>
<p>备注：对已有的 models 进行修改，Django 1.7之前的版本的Django都是无法自动更改表结构的，不过有第三方工具 south,详见 Django 数据库迁移 一节。</p>
<h1 id="使用开发服务器"><a href="#使用开发服务器" class="headerlink" title="使用开发服务器"></a>使用开发服务器</h1><p>开发服务器，即开发时使用，一般修改代码后会自动重启，方便调试和开发，但是由于性能问题，建议只用来测试，不要用在生产环境。</p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div><div class="line">10</div><div class="line">11</div></pre></td><td class="code"><pre><div class="line">python manage.py runserver</div><div class="line"> </div><div class="line"># 当提示端口被占用的时候，可以用其它端口：</div><div class="line">python manage.py runserver 8001</div><div class="line">python manage.py runserver 9999</div><div class="line">（当然也可以kill掉占用端口的进程，具体后面有讲，此处想知道的同学可查下 lsof 命令用法）</div><div class="line"> </div><div class="line"># 监听机器所有可用 ip （电脑可能有多个内网ip或多个外网ip）</div><div class="line">python manage.py runserver 0.0.0.0:8000</div><div class="line"># 如果是外网或者局域网电脑上可以用其它电脑查看开发服务器</div><div class="line"># 访问对应的 ip加端口，比如 http://172.16.20.2:8000</div></pre></td></tr></table></figure>
<h1 id="清空数据库"><a href="#清空数据库" class="headerlink" title="清空数据库"></a>清空数据库</h1><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">python manage.py flush</div></pre></td></tr></table></figure>
<p>此命令会询问是 yes 还是 no, 选择 yes 会把数据全部清空掉，只留下空表。</p>
<h1 id="创建超级管理员"><a href="#创建超级管理员" class="headerlink" title="创建超级管理员"></a>创建超级管理员</h1><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div></pre></td><td class="code"><pre><div class="line">python manage.py createsuperuser</div><div class="line"># 按照提示输入用户名和对应的密码就好了邮箱可以留空，用户名和密码必填</div><div class="line"> </div><div class="line"># 修改 用户密码可以用：</div><div class="line">python manage.py changepassword username</div></pre></td></tr></table></figure>
<h1 id="导出数据-导入数据"><a href="#导出数据-导入数据" class="headerlink" title="导出数据 导入数据"></a>导出数据 导入数据</h1><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div></pre></td><td class="code"><pre><div class="line">python manage.py dumpdata appname &gt; appname.json</div><div class="line">python manage.py loaddata appname.json</div></pre></td></tr></table></figure>
<p>关于数据操作 详见：数据导入数据迁移，现在了解有这个用法就可以了。</p>
<h1 id="Django-项目环境终端"><a href="#Django-项目环境终端" class="headerlink" title="Django 项目环境终端"></a>Django 项目环境终端</h1><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div></pre></td><td class="code"><pre><div class="line">python manage.py shell</div><div class="line">如果你安装了 bpython 或 ipython 会自动用它们的界面，推荐安装 bpython。</div></pre></td></tr></table></figure>
<p>这个命令和 直接运行 python 或 bpython 进入 shell 的区别是：你可以在这个 shell 里面调用当前项目的 models.py 中的 API，对于操作数据，还有一些小测试非常方便。</p>
<h1 id="数据库命令行"><a href="#数据库命令行" class="headerlink" title="数据库命令行"></a>数据库命令行</h1><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">python manage.py dbshell</div></pre></td></tr></table></figure>
<p>Django 会自动进入在settings.py中设置的数据库，如果是 MySQL 或 postgreSQL,会要求输入数据库用户密码。</p>
<p>在这个终端可以执行数据库的SQL语句。如果您对SQL比较熟悉，可能喜欢这种方式。</p>
<h1 id="更多命令"><a href="#更多命令" class="headerlink" title="更多命令"></a>更多命令</h1><p>终端上输入 python manage.py 可以看到详细的列表，在忘记子名称的时候特别有用。</p>
]]></content>
    
    <summary type="html">
    
      &lt;p&gt;本节主要是为了让您了解一些django最基本的命令，请尝试着记住它们，并且多多练习下，打开 Linux 或 MacOS 的 Terminal （终端）直接在 终端中输入这些命令（不是 python 的 shell中）&lt;/p&gt;
&lt;hr&gt;
    
    </summary>
    
      <category term="Django" scheme="http://feiyang.li/categories/Django/"/>
    
    
      <category term="教程" scheme="http://feiyang.li/tags/%E6%95%99%E7%A8%8B/"/>
    
      <category term="Django" scheme="http://feiyang.li/tags/Django/"/>
    
  </entry>
  
  <entry>
    <title>Coursera ML(12)-Machine Learning System Design</title>
    <link href="http://feiyang.li/2017/05/19/Coursera-ML-12-Machine-Learning-System-Design/index.html"/>
    <id>http://feiyang.li/2017/05/19/Coursera-ML-12-Machine-Learning-System-Design/index.html</id>
    <published>2017-05-19T10:53:03.000Z</published>
    <updated>2017-05-22T06:24:40.702Z</updated>
    
    <content type="html"><![CDATA[<script src="/assets/js/DPlayer.min.js"> </script><script src="/assets/js/APlayer.min.js"> </script><blockquote>
<p>机器学习系统设计</p>
</blockquote>
<hr>
<a id="more"></a>
<h1 id="查准率-Precision-和查全率-Recall"><a href="#查准率-Precision-和查全率-Recall" class="headerlink" title="查准率(Precision)和查全率(Recall)"></a>查准率(Precision)和查全率(Recall)</h1><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170519/112809620.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<p>分为四种情况：</p>
<ul>
<li>正确肯定（True Positive<strong>TP</strong>）：预测为真，实际为真。</li>
<li>正确否定（True Negative <strong>TN</strong>）：预测为假，实际为假。</li>
<li>错误肯定（False Positive<strong>FP</strong>）：预测为真，实际为假。</li>
<li>错误否定（False Negative<strong>FN</strong>）：预测为假，实际为真。<h2 id="查准率"><a href="#查准率" class="headerlink" title="查准率"></a>查准率</h2>$$Precison=\dfrac{TP}{TP+FP}$$<br>在所有我们<strong>预测有恶性肿瘤的病人</strong>中，实际上有恶性肿瘤的病人百分比，越高越好。<h2 id="查全率"><a href="#查全率" class="headerlink" title="查全率"></a>查全率</h2>$$Recall=\dfrac{TP}{TP+FN}$$<br>在所有我们<strong>实际上有病的病人</strong>中，预测成功有恶性肿瘤的病人百分比，越高越好。</li>
</ul>
<h1 id="F1值"><a href="#F1值" class="headerlink" title="F1值"></a>F1值</h1><p>$$F_1=\dfrac{PR}{P+R}$$</p>
]]></content>
    
    <summary type="html">
    
      &lt;blockquote&gt;
&lt;p&gt;机器学习系统设计&lt;/p&gt;
&lt;/blockquote&gt;
&lt;hr&gt;
    
    </summary>
    
      <category term="Coursera ML" scheme="http://feiyang.li/categories/Coursera-ML/"/>
    
    
      <category term="Machine Learning" scheme="http://feiyang.li/tags/Machine-Learning/"/>
    
      <category term="读书笔记" scheme="http://feiyang.li/tags/%E8%AF%BB%E4%B9%A6%E7%AC%94%E8%AE%B0/"/>
    
  </entry>
  
  <entry>
    <title>coursera</title>
    <link href="http://feiyang.li/2017/05/18/coursera/index.html"/>
    <id>http://feiyang.li/2017/05/18/coursera/index.html</id>
    <published>2017-05-18T02:14:53.000Z</published>
    <updated>2017-06-02T05:04:20.210Z</updated>
    
    <content type="html"><![CDATA[<script src="/assets/js/DPlayer.min.js"> </script><script src="/assets/js/APlayer.min.js"> </script><blockquote>
<p>关于coursera无法播放，原因其实很简单，就是视频所在服务器被GFW墙了。</p>
</blockquote>
<hr>
<a id="more"></a>
<p>有的时候，我真的会觉得，墙是不是太密了，使得我们不得不选择越过它。</p>
<p>常见的方法是，我们使用一个在墙外的服务器，他可以帮助我们将coursera的流量通过服务器重新转发，然后再原路返回，避过GFW。</p>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170518/101006060.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="mark" title="">
                </div>
                <div class="image-caption">mark</div>
            </figure>
<p>因此，首先要获得服务器端口以及相应的密码，这样服务器才可以帮你转发流量。我自己因为也在coursera看视频，所以自己搭建了一个服务器，分享给大家。<a href="https://ss.91vps.club/" target="_blank" rel="external">https://ss.91vps.club/</a></p>
<p>这里我分别举例展示如何快速观看coursera视频。</p>
<p>–</p>
<h1 id="Windows"><a href="#Windows" class="headerlink" title="Windows"></a>Windows</h1><h2 id="ss"><a href="#ss" class="headerlink" title="ss"></a>ss</h2><p><a href="http://home.ustc.edu.cn/~mmmwhy/Shadowsocks.exe" target="_blank" rel="external">http://home.ustc.edu.cn/~mmmwhy/Shadowsocks.exe</a></p>
<ul>
<li>在任务栏找到 Shadowsocks 图标</li>
<li>在 服务器 -&gt; 编辑服务器-&gt; 输入在网站内获得的服务器信息</li>
<li>选择 启用系统代理 来启用系统代理。请禁用浏览器里的代理插件，或把它们设置为使用系统代理。<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170515/104128278.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<h2 id="ssr"><a href="#ssr" class="headerlink" title="ssr"></a>ssr</h2><a href="http://home.ustc.edu.cn/~mmmwhy/ShadowsocksR-dotnet4.0.exe" target="_blank" rel="external">http://home.ustc.edu.cn/~mmmwhy/ShadowsocksR-dotnet4.0.exe</a><br>右键 纸飞机 -&gt; 扫描二维码 -&gt;  开启系统代理选择 -&gt;pac模式或者全局模式<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170515/104039212.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
</li>
</ul>
<p>当然直接使用二维码扫描会更方便，如果遇到某个节点不能使用，可以换一个，因为本地运营商可能会把某些ip墙掉。</p>
<h1 id="Mac-OS-X"><a href="#Mac-OS-X" class="headerlink" title="Mac OS X"></a>Mac OS X</h1><p><a href="https://github.com/shadowsocks/ShadowsocksX-NG/releases" target="_blank" rel="external">https://github.com/shadowsocks/ShadowsocksX-NG/releases</a><br>OS X 下设置使用方法<br>Mac 下使用极其方便，因为 GUI 已经集成了系统 Pac，所以无需设置任何科学上网的姿势，直接运行 ShadowsocksX ，然后输入服务商提供给你的账号信息，保存运行后，即可在所有浏览器里直接科学上网，如图：<br><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170515/105527529.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure><br><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170515/105543198.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure></p>
<h1 id="Linux"><a href="#Linux" class="headerlink" title="Linux"></a>Linux</h1><p><a href="https://github.com/shadowsocks/shadowsocks-qt5/wiki/Installation" target="_blank" rel="external">https://github.com/shadowsocks/shadowsocks-qt5/wiki/Installation</a></p>
<h1 id="Android"><a href="#Android" class="headerlink" title="Android"></a>Android</h1><p><a href="http://cdn.mmmxcc.cn/ssr_3.0.2.2.apk" target="_blank" rel="external">ssr</a><br><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170505/143305745.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure><br>扫描二维码-&gt; 然后点小飞机就可以了</p>
<h1 id="iOS"><a href="#iOS" class="headerlink" title="iOS"></a>iOS</h1><p><a href="https://itunes.apple.com/cn/app/ssrconnect-proxy-utility/id1161221960?mt=8" target="_blank" rel="external">SsrConnect - Proxy Utility</a> 免费<br><a href="https://itunes.apple.com/us/app/alice-network-proxy-utility/id1135320992?mt=8" target="_blank" rel="external">Network Proxy Utility</a> 免费<br><a href="https://itunes.apple.com/cn/app/shadowrocket/id932747118?mt=8" target="_blank" rel="external">Shadowrocket</a> 付费<br><a href="https://itunes.apple.com/app/apple-store/id1070901416?mt=8" target="_blank" rel="external">土豆丝</a> 付费</p>
]]></content>
    
    <summary type="html">
    
      &lt;blockquote&gt;
&lt;p&gt;关于coursera无法播放，原因其实很简单，就是视频所在服务器被GFW墙了。&lt;/p&gt;
&lt;/blockquote&gt;
&lt;hr&gt;
    
    </summary>
    
      <category term="shadowsocks" scheme="http://feiyang.li/categories/shadowsocks/"/>
    
    
      <category term="GFW" scheme="http://feiyang.li/tags/GFW/"/>
    
  </entry>
  
  <entry>
    <title>Coursera ML(11)-Programming Exercise 5 机器学习诊断法</title>
    <link href="http://feiyang.li/2017/05/17/Coursera-ML-11-Programming-Exercise-5/index.html"/>
    <id>http://feiyang.li/2017/05/17/Coursera-ML-11-Programming-Exercise-5/index.html</id>
    <published>2017-05-17T05:37:09.000Z</published>
    <updated>2017-05-17T06:22:10.492Z</updated>
    
    <content type="html"><![CDATA[<script src="/assets/js/DPlayer.min.js"> </script><script src="/assets/js/APlayer.min.js"> </script><blockquote>
<p>对应 <a href="http://feiyang.li/2017/05/11/Coursera-ML-10-Evaluating-a-Hypothesis/index.html">机器学习诊断法</a></p>
</blockquote>
<hr>
<a id="more"></a>
<h1 id="复习内容"><a href="#复习内容" class="headerlink" title="复习内容"></a>复习内容</h1><h2 id="Regularized-linear-regression-cost-function"><a href="#Regularized-linear-regression-cost-function" class="headerlink" title="Regularized linear regression cost function"></a>Regularized linear regression cost function</h2><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170517/104444345.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<p>对应matlab代码：<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div></pre></td><td class="code"><pre><div class="line">J = 1/(2*m)*sum((X*theta - y).^2) ;</div><div class="line">regularize =  lambda /(2*m)* (sum(theta.^2)-theta(1).^2);</div><div class="line">J = J + regularize;</div></pre></td></tr></table></figure></p>
<h2 id="Regularized-linear-regression-gradient"><a href="#Regularized-linear-regression-gradient" class="headerlink" title="Regularized linear regression gradient"></a>Regularized linear regression gradient</h2><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170517/105630413.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<p>如果对这里有疑惑，这个部分也有相应的推导过程：<br><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170325/182301586.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure><br>梯度对应代码：<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div></pre></td><td class="code"><pre><div class="line">n = size(theta);</div><div class="line">for i=1:n</div><div class="line">    grad(i) = 1/m * sum((X*theta - y).*X(:,i)) + lambda/m *theta(i);</div><div class="line">end</div><div class="line">grad(1) = 1/m * sum((X*theta - y).^X(:,1));</div></pre></td></tr></table></figure></p>
<h2 id="Learning-curves"><a href="#Learning-curves" class="headerlink" title="Learning curves"></a>Learning curves</h2><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170511/101154504.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<h2 id="Polynomial-regression"><a href="#Polynomial-regression" class="headerlink" title="Polynomial regression"></a>Polynomial regression</h2><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170517/132450490.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<h1 id="Programming-Exercise-5-Regularized-Linear-Regression-and-Bias-v-s-Variance"><a href="#Programming-Exercise-5-Regularized-Linear-Regression-and-Bias-v-s-Variance" class="headerlink" title="Programming Exercise 5 - Regularized Linear Regression and Bias v.s. Variance"></a>Programming Exercise 5 - Regularized Linear Regression and Bias v.s. Variance</h1><figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div><div class="line">10</div><div class="line">11</div><div class="line">12</div><div class="line">13</div><div class="line">14</div><div class="line">15</div><div class="line">16</div><div class="line">17</div><div class="line">18</div><div class="line">19</div><div class="line">20</div><div class="line">21</div><div class="line">22</div><div class="line">23</div></pre></td><td class="code"><pre><div class="line"><span class="comment"># %load ../../../standard_import.txt</span></div><div class="line"><span class="keyword">import</span> pandas <span class="keyword">as</span> pd</div><div class="line"><span class="keyword">import</span> numpy <span class="keyword">as</span> np</div><div class="line"><span class="keyword">import</span> matplotlib <span class="keyword">as</span> mpl</div><div class="line"><span class="keyword">import</span> matplotlib.pyplot <span class="keyword">as</span> plt</div><div class="line"></div><div class="line"><span class="keyword">from</span> scipy.io <span class="keyword">import</span> loadmat</div><div class="line"><span class="keyword">from</span> scipy.optimize <span class="keyword">import</span> minimize</div><div class="line"></div><div class="line"><span class="keyword">from</span> sklearn.linear_model <span class="keyword">import</span> LinearRegression, Ridge</div><div class="line"><span class="keyword">from</span> sklearn.preprocessing <span class="keyword">import</span> PolynomialFeatures</div><div class="line"></div><div class="line">pd.set_option(<span class="string">'display.notebook_repr_html'</span>, <span class="keyword">False</span>)</div><div class="line">pd.set_option(<span class="string">'display.max_columns'</span>, <span class="keyword">None</span>)</div><div class="line">pd.set_option(<span class="string">'display.max_rows'</span>, <span class="number">150</span>)</div><div class="line">pd.set_option(<span class="string">'display.max_seq_items'</span>, <span class="keyword">None</span>)</div><div class="line"> </div><div class="line"><span class="comment">#%config InlineBackend.figure_formats = &#123;'pdf',&#125;</span></div><div class="line">%matplotlib inline</div><div class="line"></div><div class="line"><span class="keyword">import</span> seaborn <span class="keyword">as</span> sns</div><div class="line">sns.set_context(<span class="string">'notebook'</span>)</div><div class="line">sns.set_style(<span class="string">'white'</span>)</div></pre></td></tr></table></figure>
<figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div></pre></td><td class="code"><pre><div class="line">data = loadmat(<span class="string">'data/ex5data1.mat'</span>)</div><div class="line">data.keys()</div></pre></td></tr></table></figure>
<pre><code>dict_keys([&apos;__header__&apos;, &apos;__version__&apos;, &apos;__globals__&apos;, &apos;X&apos;, &apos;y&apos;, &apos;Xtest&apos;, &apos;ytest&apos;, &apos;Xval&apos;, &apos;yval&apos;])
</code></pre><figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div><div class="line">10</div><div class="line">11</div></pre></td><td class="code"><pre><div class="line">y_train = data[<span class="string">'y'</span>]</div><div class="line">X_train = np.c_[np.ones_like(data[<span class="string">'X'</span>]), data[<span class="string">'X'</span>]]</div><div class="line"></div><div class="line">yval = data[<span class="string">'yval'</span>]</div><div class="line">Xval = np.c_[np.ones_like(data[<span class="string">'Xval'</span>]), data[<span class="string">'Xval'</span>]]</div><div class="line"></div><div class="line"></div><div class="line">print(<span class="string">'X_train:'</span>, X_train.shape)</div><div class="line">print(<span class="string">'y_train:'</span>, y_train.shape)</div><div class="line">print(<span class="string">'Xval:'</span>, Xval.shape)</div><div class="line">print(<span class="string">'yval:'</span>, yval.shape)</div></pre></td></tr></table></figure>
<pre><code>X_train: (12, 2)
y_train: (12, 1)
Xval: (21, 2)
yval: (21, 1)
</code></pre><h2 id="Regularized-Linear-Regression"><a href="#Regularized-Linear-Regression" class="headerlink" title="Regularized Linear Regression"></a>Regularized Linear Regression</h2><figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div></pre></td><td class="code"><pre><div class="line">plt.scatter(X_train[:,<span class="number">1</span>], y_train, s=<span class="number">50</span>, c=<span class="string">'r'</span>, marker=<span class="string">'x'</span>, linewidths=<span class="number">1</span>)</div><div class="line">plt.xlabel(<span class="string">'Change in water level (x)'</span>)</div><div class="line">plt.ylabel(<span class="string">'Water flowing out of the dam (y)'</span>)</div><div class="line">plt.ylim(ymin=<span class="number">0</span>);</div></pre></td></tr></table></figure>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170517/133552305.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<p>###Regularized Cost function</p>
<figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div></pre></td><td class="code"><pre><div class="line"><span class="function"><span class="keyword">def</span> <span class="title">linearRegCostFunction</span><span class="params">(theta, X, y, reg)</span>:</span></div><div class="line">    m = y.size</div><div class="line">    </div><div class="line">    h = X.dot(theta)</div><div class="line">    </div><div class="line">    J = (<span class="number">1</span>/(<span class="number">2</span>*m))*np.sum(np.square(h-y)) + (reg/(<span class="number">2</span>*m))*np.sum(np.square(theta[<span class="number">1</span>:]))</div><div class="line">   </div><div class="line">    <span class="keyword">return</span>(J)</div></pre></td></tr></table></figure>
<h3 id="Gradient"><a href="#Gradient" class="headerlink" title="Gradient"></a>Gradient</h3><figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div></pre></td><td class="code"><pre><div class="line"><span class="function"><span class="keyword">def</span> <span class="title">lrgradientReg</span><span class="params">(theta, X, y, reg)</span>:</span></div><div class="line">    m = y.size</div><div class="line">    </div><div class="line">    h = X.dot(theta.reshape(<span class="number">-1</span>,<span class="number">1</span>))</div><div class="line">        </div><div class="line">    grad = (<span class="number">1</span>/m)*(X.T.dot(h-y))+ (reg/m)*np.r_[[[<span class="number">0</span>]],theta[<span class="number">1</span>:].reshape(<span class="number">-1</span>,<span class="number">1</span>)]</div><div class="line">        </div><div class="line">    <span class="keyword">return</span>(grad.flatten())</div></pre></td></tr></table></figure>
<figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div></pre></td><td class="code"><pre><div class="line">initial_theta = np.ones((X_train.shape[<span class="number">1</span>],<span class="number">1</span>))</div><div class="line">cost = linearRegCostFunction(initial_theta, X_train, y_train, <span class="number">0</span>)</div><div class="line">gradient = lrgradientReg(initial_theta, X_train, y_train, <span class="number">0</span>)</div><div class="line">print(cost)</div><div class="line">print(gradient)</div></pre></td></tr></table></figure>
<pre><code>303.951525554
[ -15.30301567  598.16741084]
</code></pre><figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div><div class="line">10</div></pre></td><td class="code"><pre><div class="line"><span class="function"><span class="keyword">def</span> <span class="title">trainLinearReg</span><span class="params">(X, y, reg)</span>:</span></div><div class="line">    <span class="comment">#initial_theta = np.zeros((X.shape[1],1))</span></div><div class="line">    initial_theta = np.array([[<span class="number">15</span>],[<span class="number">15</span>]])</div><div class="line">    <span class="comment"># For some reason the minimize() function does not converge when using</span></div><div class="line">    <span class="comment"># zeros as initial theta.</span></div><div class="line">        </div><div class="line">    res = minimize(linearRegCostFunction, initial_theta, args=(X,y,reg), method=<span class="keyword">None</span>, jac=lrgradientReg,</div><div class="line">                   options=&#123;<span class="string">'maxiter'</span>:<span class="number">5000</span>&#125;)</div><div class="line">    </div><div class="line">    <span class="keyword">return</span>(res)</div></pre></td></tr></table></figure>
<figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div></pre></td><td class="code"><pre><div class="line">fit = trainLinearReg(X_train, y_train, <span class="number">0</span>)</div><div class="line">fit</div></pre></td></tr></table></figure>
<pre><code>     fun: 1604.4002999186634
hess_inv: array([[ 1.03142187,  0.00617881],
      [ 0.00617881,  0.001215  ]])
     jac: array([  3.42437190e-12,  -5.70371898e-10])
 message: &apos;Optimization terminated successfully.&apos;
    nfev: 6
     nit: 4
    njev: 6
  status: 0
 success: True
       x: array([ 13.08790351,   0.36777923])
</code></pre><h3 id="Comparison-coefficients-and-cost-obtained-with-LinearRegression-in-Scikit-learn"><a href="#Comparison-coefficients-and-cost-obtained-with-LinearRegression-in-Scikit-learn" class="headerlink" title="Comparison: coefficients and cost obtained with LinearRegression in Scikit-learn"></a>Comparison: coefficients and cost obtained with LinearRegression in Scikit-learn</h3><figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div></pre></td><td class="code"><pre><div class="line">regr = LinearRegression(fit_intercept=<span class="keyword">False</span>)</div><div class="line">regr.fit(X_train, y_train.ravel())</div><div class="line">print(regr.coef_)</div><div class="line">print(linearRegCostFunction(regr.coef_, X_train, y_train, <span class="number">0</span>))</div></pre></td></tr></table></figure>
<pre><code>[ 13.08790351   0.36777923]
1604.40029992
</code></pre><figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div></pre></td><td class="code"><pre><div class="line">plt.plot(np.linspace(<span class="number">-50</span>,<span class="number">40</span>), (fit.x[<span class="number">0</span>]+ (fit.x[<span class="number">1</span>]*np.linspace(<span class="number">-50</span>,<span class="number">40</span>))), label=<span class="string">'Scipy optimize'</span>)</div><div class="line"><span class="comment">#plt.plot(np.linspace(-50,40), (regr.coef_[0]+ (regr.coef_[1]*np.linspace(-50,40))), label='Scikit-learn')</span></div><div class="line">plt.scatter(X_train[:,<span class="number">1</span>], y_train, s=<span class="number">50</span>, c=<span class="string">'r'</span>, marker=<span class="string">'x'</span>, linewidths=<span class="number">1</span>)</div><div class="line">plt.xlabel(<span class="string">'Change in water level (x)'</span>)</div><div class="line">plt.ylabel(<span class="string">'Water flowing out of the dam (y)'</span>)</div><div class="line">plt.ylim(ymin=<span class="number">-5</span>)</div><div class="line">plt.xlim(xmin=<span class="number">-50</span>)</div><div class="line">plt.legend(loc=<span class="number">4</span>);</div></pre></td></tr></table></figure>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170517/133621870.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div><div class="line">10</div><div class="line">11</div><div class="line">12</div></pre></td><td class="code"><pre><div class="line"><span class="function"><span class="keyword">def</span> <span class="title">learningCurve</span><span class="params">(X, y, Xval, yval, reg)</span>:</span></div><div class="line">    m = y.size</div><div class="line">    </div><div class="line">    error_train = np.zeros((m, <span class="number">1</span>))</div><div class="line">    error_val = np.zeros((m, <span class="number">1</span>))</div><div class="line">    </div><div class="line">    <span class="keyword">for</span> i <span class="keyword">in</span> np.arange(m):</div><div class="line">        res = trainLinearReg(X[:i+<span class="number">1</span>], y[:i+<span class="number">1</span>], reg)</div><div class="line">        error_train[i] = linearRegCostFunction(res.x, X[:i+<span class="number">1</span>], y[:i+<span class="number">1</span>], reg)</div><div class="line">        error_val[i] = linearRegCostFunction(res.x, Xval, yval, reg)</div><div class="line">    </div><div class="line">    <span class="keyword">return</span>(error_train, error_val)</div></pre></td></tr></table></figure>
<figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">t_error, v_error = learningCurve(X_train, y_train, Xval, yval, <span class="number">0</span>)</div></pre></td></tr></table></figure>
<figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div></pre></td><td class="code"><pre><div class="line">plt.plot(np.arange(<span class="number">1</span>,<span class="number">13</span>), t_error, label=<span class="string">'Training error'</span>)</div><div class="line">plt.plot(np.arange(<span class="number">1</span>,<span class="number">13</span>), v_error, label=<span class="string">'Validation error'</span>)</div><div class="line">plt.title(<span class="string">'Learning curve for linear regression'</span>)</div><div class="line">plt.xlabel(<span class="string">'Number of training examples'</span>)</div><div class="line">plt.ylabel(<span class="string">'Error'</span>)</div><div class="line">plt.legend();</div></pre></td></tr></table></figure>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170517/133630516.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<h3 id="Polynomial-regression-Scikit-learn"><a href="#Polynomial-regression-Scikit-learn" class="headerlink" title="Polynomial regression (Scikit-learn)"></a>Polynomial regression (Scikit-learn)</h3><figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div><div class="line">10</div><div class="line">11</div><div class="line">12</div><div class="line">13</div><div class="line">14</div><div class="line">15</div><div class="line">16</div><div class="line">17</div><div class="line">18</div><div class="line">19</div><div class="line">20</div><div class="line">21</div><div class="line">22</div></pre></td><td class="code"><pre><div class="line">poly = PolynomialFeatures(degree=<span class="number">8</span>)</div><div class="line">X_train_poly = poly.fit_transform(X_train[:,<span class="number">1</span>].reshape(<span class="number">-1</span>,<span class="number">1</span>))</div><div class="line"></div><div class="line">regr2 = LinearRegression()</div><div class="line">regr2.fit(X_train_poly, y_train)</div><div class="line"></div><div class="line">regr3 = Ridge(alpha=<span class="number">20</span>)</div><div class="line">regr3.fit(X_train_poly, y_train)</div><div class="line"></div><div class="line"><span class="comment"># plot range for x</span></div><div class="line">plot_x = np.linspace(<span class="number">-60</span>,<span class="number">45</span>)</div><div class="line"><span class="comment"># using coefficients to calculate y</span></div><div class="line">plot_y = regr2.intercept_+ np.sum(regr2.coef_*poly.fit_transform(plot_x.reshape(<span class="number">-1</span>,<span class="number">1</span>)), axis=<span class="number">1</span>)</div><div class="line">plot_y2 = regr3.intercept_ + np.sum(regr3.coef_*poly.fit_transform(plot_x.reshape(<span class="number">-1</span>,<span class="number">1</span>)), axis=<span class="number">1</span>)</div><div class="line"></div><div class="line">plt.plot(plot_x, plot_y, label=<span class="string">'Scikit-learn LinearRegression'</span>)</div><div class="line">plt.plot(plot_x, plot_y2, label=<span class="string">'Scikit-learn Ridge (alpha=&#123;&#125;)'</span>.format(regr3.alpha))</div><div class="line">plt.scatter(X_train[:,<span class="number">1</span>], y_train, s=<span class="number">50</span>, c=<span class="string">'r'</span>, marker=<span class="string">'x'</span>, linewidths=<span class="number">1</span>)</div><div class="line">plt.xlabel(<span class="string">'Change in water level (x)'</span>)</div><div class="line">plt.ylabel(<span class="string">'Water flowing out of the dam (y)'</span>)</div><div class="line">plt.title(<span class="string">'Polynomial regression degree 8'</span>)</div><div class="line">plt.legend(loc=<span class="number">4</span>);</div></pre></td></tr></table></figure>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170517/133642078.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
]]></content>
    
    <summary type="html">
    
      &lt;blockquote&gt;
&lt;p&gt;对应 &lt;a href=&quot;http://feiyang.li/2017/05/11/Coursera-ML-10-Evaluating-a-Hypothesis/index.html&quot;&gt;机器学习诊断法&lt;/a&gt;&lt;/p&gt;
&lt;/blockquote&gt;
&lt;hr&gt;
    
    </summary>
    
      <category term="Coursera ML" scheme="http://feiyang.li/categories/Coursera-ML/"/>
    
    
      <category term="Machine Learning" scheme="http://feiyang.li/tags/Machine-Learning/"/>
    
      <category term="读书笔记" scheme="http://feiyang.li/tags/%E8%AF%BB%E4%B9%A6%E7%AC%94%E8%AE%B0/"/>
    
      <category term="Python" scheme="http://feiyang.li/tags/Python/"/>
    
  </entry>
  
  <entry>
    <title>数据库操作</title>
    <link href="http://feiyang.li/2017/05/16/databases/index.html"/>
    <id>http://feiyang.li/2017/05/16/databases/index.html</id>
    <published>2017-05-16T10:38:00.000Z</published>
    <updated>2017-05-31T13:00:54.295Z</updated>
    
    <content type="html"><![CDATA[<script src="/assets/js/DPlayer.min.js"> </script><script src="/assets/js/APlayer.min.js"> </script><blockquote>
<p>其实之前学过数据库相关操作，只不过许久不用就忘了。真是年纪大了呀。</p>
</blockquote>
<hr>
<a id="more"></a>
<ul>
<li><p>创建数据库：<code>create database books;</code></p>
</li>
<li><p>使用数据库：<code>use dbname;</code></p>
</li>
</ul>
<ul>
<li>新建数据库表<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div></pre></td><td class="code"><pre><div class="line">create table tablename(columns);</div><div class="line">create table demo(</div><div class="line">     userid int unsigned not null auto_increment primary key,</div><div class="line">     username char(50) not null,</div><div class="line">     password char(20) not null,</div><div class="line">     age int not null,</div><div class="line">     city char(20) not null     </div><div class="line">);</div></pre></td></tr></table></figure>
</li>
</ul>
<ul>
<li><p>显示数据库表：<code>show tables;</code></p>
</li>
<li><p>在数据库中插入数据</p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div></pre></td><td class="code"><pre><div class="line">INSERT INTO `jobcategory` VALUES(11,&quot;丸剂&quot;);</div><div class="line">INSERT INTO wan (JobCategoryID,JobCategoryName) SELECT ID,MingCheng FROM `zysjyj` WHERE MingCheng LIKE  &quot;%丸&quot;;</div></pre></td></tr></table></figure>
</li>
<li><p>查询数据</p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">SELECT ID,MingCheng FROM `zysjyj` WHERE MingCheng LIKE  &quot;%丸&quot;;</div></pre></td></tr></table></figure>
</li>
<li><p>更新数据：</p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">UPDATE `zysjyj` SET ID=(1100000+ID) WHERE MingCheng LIKE &apos;%丸&apos;;</div></pre></td></tr></table></figure>
</li>
<li><p>删除数据：</p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">DELETE FROM `nzysys`.`jobintroduction`;</div></pre></td></tr></table></figure>
</li>
</ul>
<ul>
<li>表的删除：<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">DROP TABLE table;</div></pre></td></tr></table></figure>
</li>
</ul>
<ul>
<li><p>数据删除：</p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">DROP DATABASE database;`</div></pre></td></tr></table></figure>
</li>
<li><p>通过<strong>SELECT</strong>来<strong>INSERT</strong>：<br>先创建新表，然后插入</p>
</li>
</ul>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div></pre></td><td class="code"><pre><div class="line">CREATE TABLE wan(JobCategoryID INT(20) NOT NULL AUTO_INCREMENT,PRIMARY KEY (JobCategoryID),parentID INT(20) NOT NULL DEFAULT &apos;10&apos; ,JobCategoryName VARCHAR(128)  NOT NULL);  </div><div class="line"></div><div class="line">INSERT INTO wan (JobCategoryID,JobCategoryName) SELECT ID,MingCheng FROM `zysjyj` WHERE MingCheng LIKE  &quot;%丸&quot;;</div></pre></td></tr></table></figure>
<ul>
<li><p>跨数据库插入</p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">INSERT `nzysys`.`subjobcategory` SELECT * FROM `chinese-drug`.`wan`;</div></pre></td></tr></table></figure>
</li>
<li><p>用<code>select</code>来<code>create</code>，并指定默认值</p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">INSERT INTO `nzysys`.`subjobcategory`(JobCategoryID,parentID,JobCategoryName)SELECT ID,10,MingCheng FROM `chinese-drug`.`zysjyj` WHERE MingCheng LIKE &quot;%散&quot;;</div></pre></td></tr></table></figure>
</li>
<li><p>叠加使用</p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div></pre></td><td class="code"><pre><div class="line">SELECT ID,MingCheng FROM `chinese-drug`.`zysjyj` WHERE MingCheng LIKE &quot;%散&quot;;</div><div class="line">UPDATE `chinese-drug`.`zysjyj` SET ID=(1000000+ID) WHERE MingCheng LIKE &quot;%散&quot;;</div><div class="line">INSERT INTO `nzysys`.`subjobcategory`(JobCategoryID,parentID,JobCategoryName)SELECT ID,10,MingCheng FROM `chinese-drug`.`zysjyj` WHERE MingCheng LIKE &quot;%散&quot;;</div></pre></td></tr></table></figure>
</li>
<li><p>自增ID归零 </p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">TRUNCATE TABLE `nzysys`.`jobintroduction`;</div></pre></td></tr></table></figure>
</li>
<li><p>交叉</p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">INSERT INTO `nzysys`.`jobintroduction` (jobCategoryID,introductionValue,introductionkey,introductionID) SELECT b.jobCategoryID,a.ChuFang,&apos;组成&apos;,1 FROM `nzysys`.`subjobcategory` AS b,`chinese-drug`.`zysjyj` AS a WHERE a.`ID`=b.`JobCategoryID`;</div></pre></td></tr></table></figure>
</li>
<li><p>字符串分割</p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">SELECT SUBSTRING_INDEX(SUBSTRING_INDEX(o.NeiRong, &apos;【&apos;, 2),&apos;】&apos;,-1) FROM `nzysys`.`drug_history` AS o;</div></pre></td></tr></table></figure>
</li>
</ul>
<p>之前都是一个字符串一个字符串分割，这个操作可以把整个mysql某个列全部分割，洋气</p>
<ul>
<li>分割后插入<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div></pre></td><td class="code"><pre><div class="line">INSERT INTO `nzysys`.`medical_history`(subMajorMLID,subMajorMLName,subMajorTitle,subMajorDetail) </div><div class="line">    SELECT ID,BiaoTi,</div><div class="line">    SUBSTRING_INDEX(SUBSTRING_INDEX(o.NeiRong, &apos;【&apos;, 2),&apos;】&apos;,-1),</div><div class="line">    SUBSTRING_INDEX(SUBSTRING_INDEX(o.NeiRong, &apos;【&apos;, 5),&apos;【病因】&apos;,-1)FROM `nzysys`.`drug_history` AS o;</div></pre></td></tr></table></figure></li>
</ul>
]]></content>
    
    <summary type="html">
    
      &lt;blockquote&gt;
&lt;p&gt;其实之前学过数据库相关操作，只不过许久不用就忘了。真是年纪大了呀。&lt;/p&gt;
&lt;/blockquote&gt;
&lt;hr&gt;
    
    </summary>
    
    
      <category term="databases" scheme="http://feiyang.li/tags/databases/"/>
    
  </entry>
  
  <entry>
    <title>免费vps get!</title>
    <link href="http://feiyang.li/2017/05/15/free-vps/index.html"/>
    <id>http://feiyang.li/2017/05/15/free-vps/index.html</id>
    <published>2017-05-15T12:13:05.000Z</published>
    <updated>2017-05-15T12:13:29.549Z</updated>
    
    <content type="html"><![CDATA[<script src="/assets/js/DPlayer.min.js"> </script><script src="/assets/js/APlayer.min.js"> </script><blockquote>
<p>早就听说WisHosting 长期免费vps，码农买vps就像女人买衣服一样，虽然没什么用，但是会上瘾。</p>
</blockquote>
<hr>
<a id="more"></a>
<h1 id="要求"><a href="#要求" class="headerlink" title="要求"></a>要求</h1><p><a href="https://www.wishosting.com/plugin/support_manager/knowledgebase/view/14/how-to-get-openvz-mini-for-free/" target="_blank" rel="external">官网链接</a></p>
<p>How to get OpenVZ Mini for free?<br>If you want to grab free VPS (OpenVZ Mini package) free of charge, here is the list of requirements you should meet:<br>1) The domain from <strong>non-free domain zone</strong> (for example .com)<br>2) The link to <a href="https://www.wishosting.com/" target="_blank" rel="external">https://www.wishosting.com/</a> on your domain index page.<br>3) Your website should have some real content (5+ pages) and it should be indexed by search engines.<br>4) The service will be active as long as you keep the backlink on your website.</p>
<p>Please open a support ticket if you meet those requirements or send an e-mail to support@wishosting.com</p>
<h1 id="动手"><a href="#动手" class="headerlink" title="动手"></a>动手</h1><ul>
<li><p>注册并下单，不需要支付</p>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170515/195827250.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
</li>
<li><p>发工单询问商家，按照沟通步骤来即可</p>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170515/200957854.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
</li>
</ul>
<hr>
<p>表示已经申请到了，所以，应该不是很难申请的，有兴趣可以去玩一下。</p>
]]></content>
    
    <summary type="html">
    
      &lt;blockquote&gt;
&lt;p&gt;早就听说WisHosting 长期免费vps，码农买vps就像女人买衣服一样，虽然没什么用，但是会上瘾。&lt;/p&gt;
&lt;/blockquote&gt;
&lt;hr&gt;
    
    </summary>
    
    
      <category term="vps" scheme="http://feiyang.li/tags/vps/"/>
    
      <category term="free" scheme="http://feiyang.li/tags/free/"/>
    
  </entry>
  
  <entry>
    <title>shadowsocks-manager一键安装脚本</title>
    <link href="http://feiyang.li/2017/05/14/ss-mgr/index.html"/>
    <id>http://feiyang.li/2017/05/14/ss-mgr/index.html</id>
    <published>2017-05-14T03:58:57.000Z</published>
    <updated>2017-05-30T02:01:43.723Z</updated>
    
    <content type="html"><![CDATA[<script src="/assets/js/DPlayer.min.js"> </script><script src="/assets/js/APlayer.min.js"> </script><blockquote>
<p>shadowsocks-manager是一个非常帅气的ss多用户管理程序，搭建起来稍微有点麻烦，因此写了本一键脚本。示例网站：<a href="http://gz.feiyang.li/" target="_blank" rel="external">http://gz.feiyang.li/</a></p>
</blockquote>
<hr>
<a id="more"></a>
<p>我看了一下数据，大概70%的人会一次性安装成功，其余人大多卡在npm依赖安装不全的问题上；<br>请注意：<br>1、内存1g以下的vps，极有可能内存过小，导致依赖被kill掉。<br>2、网络问题，请多次安装npm。<br>如果实在解决不了，建议就不要用这个管理界面，这个管理界面的要求的确有点高，请理解。</p>
<h1 id="效果"><a href="#效果" class="headerlink" title="效果"></a>效果</h1><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170513/135344468.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170513/135349156.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170513/135357497.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170513/140400232.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<p>github： <a href="https://github.com/mmmwhy/ss-mgr" target="_blank" rel="external">https://github.com/mmmwhy/ss-mgr</a></p>
<h1 id="特点"><a href="#特点" class="headerlink" title="特点"></a>特点</h1><ul>
<li>非常帅气，各种图表一上，是不是很牛逼的感觉。</li>
<li>支持支付宝付款对账，站长们多出来的vps可以不用再继续吃灰了。</li>
</ul>
<h1 id="要求"><a href="#要求" class="headerlink" title="要求"></a>要求</h1><p>centos7 X64，在腾讯云，digitialocean,interserver,ethernetservers通过测试。<br>其他版本还没有测试，理论上可用。</p>
<h1 id="安装脚本"><a href="#安装脚本" class="headerlink" title="安装脚本"></a>安装脚本</h1><h2 id="安装ss-mgr"><a href="#安装ss-mgr" class="headerlink" title="安装ss-mgr"></a>安装ss-mgr</h2><p>本脚本包括主控端和节点端，安装时，自动添加本vps作为一个节点。<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">wget -N --no-check-certificate https://raw.githubusercontent.com/mmmwhy/ss-mgr/master/sm.sh &amp;&amp; chmod +x sm.sh &amp;&amp; bash sm.sh</div></pre></td></tr></table></figure></p>
<p>坐等就可以了，没有什么需要做的东西。<br>打开ip地址，就可以看到ss-mgr了。</p>
<h2 id="增加节点"><a href="#增加节点" class="headerlink" title="增加节点"></a>增加节点</h2><p>本脚本为：已有<strong>vps安装好主控端</strong>，添加新节点时，在<strong>新节点</strong>使用的脚本。<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">wget -N --no-check-certificate https://raw.githubusercontent.com/mmmwhy/ss-mgr/master/sm_node.sh &amp;&amp; chmod +x sm_node.sh &amp;&amp; bash sm_node.sh</div></pre></td></tr></table></figure></p>
<ul>
<li>输入密码<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170514/135830856.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure></li>
<li>前端页面填入，刚才的密码，选择加密方式。<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170514/140131877.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
</li>
</ul>
<h1 id="docker方法"><a href="#docker方法" class="headerlink" title="docker方法"></a>docker方法</h1><h1 id="备注"><a href="#备注" class="headerlink" title="备注"></a>备注</h1><ul>
<li>注册的第一个用户就是管理员.</li>
<li>注册第二个账号，可以看到自己的剩余时间，一般来说是8个小时。可以点击续费，选择任意一个额度，支付后，可以看到时间会增长。</li>
<li>默认使用hotmail邮箱，因为邮箱在国外，所以如果服务器在国内的话，可能会timeout。</li>
<li>某些系统内核版本过老，因此可能会出现如下症状，按回车就可以了。<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170513/135239354.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure></li>
<li>被墙掉的资源都换成国内的了。</li>
<li>三条重要命令<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div></pre></td><td class="code"><pre><div class="line">screen -dmS ss-manager ss-manager -m aes-256-cfb -u --manager-address 127.0.0.1:4000</div><div class="line">cd /root/shadowsocks-manager/</div><div class="line">screen -dmS ss node server.js -c /root/.ssmgr/ss.yml</div><div class="line">screen -dmS webgui node server.js -c /root/.ssmgr/webgui.yml</div></pre></td></tr></table></figure>
</li>
</ul>
<h2 id="修改页面"><a href="#修改页面" class="headerlink" title="修改页面"></a>修改页面</h2><p>修改过文本后，使用<code>npm run build</code>重新生成页面。<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">screen -dmS webgui node server.js -c /root/.ssmgr/webgui.yml</div></pre></td></tr></table></figure></p>
<h2 id="一键BBR脚本"><a href="#一键BBR脚本" class="headerlink" title="一键BBR脚本"></a>一键BBR脚本</h2><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div></pre></td><td class="code"><pre><div class="line">wget --no-check-certificate https://github.com/teddysun/across/raw/master/bbr.sh</div><div class="line">chmod +x bbr.sh</div><div class="line">./bbr.sh</div></pre></td></tr></table></figure>
<h2 id="增加虚拟内存"><a href="#增加虚拟内存" class="headerlink" title="增加虚拟内存"></a>增加虚拟内存</h2><p><strong>增加swap交换文件</strong></p>
<ul>
<li>使用dd命令创建一个swap交换文件<br><code>dd if=/dev/zero of=/home/swap bs=1024 count=1024000</code><br>这样就建立一个/home/swap的分区文件，大小为1G。</li>
<li>制作为swap格式文件：<br><code>mkswap /home/swap</code></li>
<li>再用swapon命令把这个文件分区挂载swap分区<br><code>/sbin/swapon /home/swap</code><br>我们用free -m命令看一下，发现已经有交换分区了。但是重启系统后，swap分区又变成0了。</li>
<li>为防止重启后swap分区变成0，要修改/etc/fstab文件<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">vi /etc/fstab</div></pre></td></tr></table></figure>
</li>
</ul>
<p>在文件末尾（最后一行）加上<br><code>/home/swap swap swap default 0 0</code><br>这样就算重启系统，swap分区还是有值。</p>
<ul>
<li>查看空间：<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">free -m</div></pre></td></tr></table></figure>
</li>
</ul>
<hr>
<p>因为本工具较难配置，因此如果有问题的话，可以通过tg联系我。<a href="https://t.me/feiyangss" target="_blank" rel="external">https://t.me/feiyangss</a></p>
<hr>
<p>参考以下链接：<br><a href="http://blog.csdn.net/lengyue1084/article/details/51405640" target="_blank" rel="external">http://blog.csdn.net/lengyue1084/article/details/51405640</a><br><a href="https://code.momok.xyz/server/deploy-ss-manager.html" target="_blank" rel="external">https://code.momok.xyz/server/deploy-ss-manager.html</a><br><a href="https://github.com/shadowsocks/shadowsocks-manager" target="_blank" rel="external">https://github.com/shadowsocks/shadowsocks-manager</a></p>
]]></content>
    
    <summary type="html">
    
      &lt;blockquote&gt;
&lt;p&gt;shadowsocks-manager是一个非常帅气的ss多用户管理程序，搭建起来稍微有点麻烦，因此写了本一键脚本。示例网站：&lt;a href=&quot;http://gz.feiyang.li/&quot;&gt;http://gz.feiyang.li/&lt;/a&gt;&lt;/p&gt;
&lt;/blockquote&gt;
&lt;hr&gt;
    
    </summary>
    
      <category term="shadowsocks" scheme="http://feiyang.li/categories/shadowsocks/"/>
    
    
      <category term="GFW" scheme="http://feiyang.li/tags/GFW/"/>
    
      <category term="Linux" scheme="http://feiyang.li/tags/Linux/"/>
    
      <category term="教程" scheme="http://feiyang.li/tags/%E6%95%99%E7%A8%8B/"/>
    
      <category term="vps" scheme="http://feiyang.li/tags/vps/"/>
    
      <category term="shell" scheme="http://feiyang.li/tags/shell/"/>
    
  </entry>
  
  <entry>
    <title>Coursera ML(10)-机器学习诊断法</title>
    <link href="http://feiyang.li/2017/05/11/Coursera-ML-10-Evaluating-a-Hypothesis/index.html"/>
    <id>http://feiyang.li/2017/05/11/Coursera-ML-10-Evaluating-a-Hypothesis/index.html</id>
    <published>2017-05-11T04:28:16.000Z</published>
    <updated>2017-05-11T04:34:51.020Z</updated>
    
    <content type="html"><![CDATA[<script src="/assets/js/DPlayer.min.js"> </script><script src="/assets/js/APlayer.min.js"> </script><blockquote>
<p>假设你在开发一个机器学习系统，或者在改进一个机器学习系统的性能，应如何做？</p>
</blockquote>
<hr>
<a id="more"></a>
<h1 id="目前已有的方法："><a href="#目前已有的方法：" class="headerlink" title="目前已有的方法："></a>目前已有的方法：</h1><ul>
<li>Getting more training examples</li>
<li>Trying smaller sets of features</li>
<li>Trying additional features</li>
<li>Trying polynomial features</li>
<li>Increasing or decreasing λ</li>
</ul>
<p>每种方法都有自己不同的应用场景</p>
<h1 id="Evaluating-a-Hypothesis"><a href="#Evaluating-a-Hypothesis" class="headerlink" title="Evaluating a Hypothesis"></a>Evaluating a Hypothesis</h1><p>根据测试集得到参数，对训练集运用模型。有两种误差计算方法</p>
<ul>
<li><p>For linear regression: </p>
 $$J_{test}(\Theta) = \dfrac{1}{2m_{test}} \sum_{i=1}^{m_{test}}(h_\Theta(x^{(i)}_{test}) - y^{(i)}_{test})^2$$ 
</li>
<li><p>For classification :<br>误分类的比例，对于每一个测试实例，计算：</p>
 $$err(h_\Theta(x),y) = \begin{matrix} 1 & \mbox{if } h_\Theta(x) \geq 0.5\ and\ y = 0\ or\ h_\Theta(x) < 0.5\ and\ y = 1\newline 0 & \mbox otherwise \end{matrix}$$
然后急死俺平均
$$\text{Test Error} = \dfrac{1}{m_{test}} \sum^{m_{test}}_{i=1} err(h_\Theta(x^{(i)}_{test}), y^{(i)}_{test})$$ 
</li>
</ul>
<h1 id="Model-Selection-and-Train-Validation-Test-Sets-交叉验证机"><a href="#Model-Selection-and-Train-Validation-Test-Sets-交叉验证机" class="headerlink" title="Model Selection and Train/Validation/Test Sets(交叉验证机)"></a>Model Selection and Train/Validation/Test Sets(交叉验证机)</h1><p>使用60%的数据作为训练集，20%的数据作为交叉验证集，20%的数据作为测试集<br><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170511/101002408.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure></p>
<ul>
<li>Optimize the parameters in Θ using the training set for each polynomial degree.</li>
<li>Find the polynomial degree d with the least error using the cross validation set.</li>
<li>Estimate the generalization error using the test set with $J_{test}(\Theta^{(d)})$, (d = theta from polynomial with lower error);<br>简单来讲：<br>训练集训练出 10 个模型 -&gt;10 个模型分别对交叉验证集计算得出交叉验证误差（代价函数的值）-&gt;选取代价函数值最小的模型-&gt;用选出的模型对测试集计算得出推广误差（代价函数的值）<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170511/101154504.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
</li>
</ul>
<h1 id="Diagnosing-Bias-vs-Variance"><a href="#Diagnosing-Bias-vs-Variance" class="headerlink" title="Diagnosing Bias vs. Variance"></a>Diagnosing Bias vs. Variance</h1><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170511/103200427.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="high biais and high variance" title="">
                </div>
                <div class="image-caption">high biais and high variance</div>
            </figure>
<ul>
<li> High bias (underfitting): both $J_{train}(\Theta)$ and $J_{CV}(\Theta)$ will be high. Also, $J_{CV}(\Theta) \approx J_{train}(\Theta)$. </li>
<li> High variance (overfitting): $J_{train}(\Theta)$ will be low and $J_{CV}(\Theta)$ will be much greater than $J_{train}(\Theta)$. 
</li>
</ul>
<p>很多情况下，欠拟合会导致高误差，高方差意味着拟合过度。</p>
<h2 id="Decide-Bias-or-Variance"><a href="#Decide-Bias-or-Variance" class="headerlink" title="Decide Bias or Variance"></a>Decide Bias or Variance</h2><ul>
<li>训练集误差和交叉验证集误差近似时：偏差/欠拟合</li>
<li>交叉训练集误差 &gt;&gt; 训练集误差时：方法/过拟合</li>
</ul>
<h1 id="Regularization-and-Bias-Variance"><a href="#Regularization-and-Bias-Variance" class="headerlink" title="Regularization and Bias/Variance"></a>Regularization and Bias/Variance</h1><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170511/110947313.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="mark" title="">
                </div>
                <div class="image-caption">mark</div>
            </figure>
<ul>
<li>Create a list of lambdas (i.e. λ∈{0,0.01,0.02,0.04,0.08,0.16,0.32,0.64,1.28,2.56,5.12,10.24});</li>
<li>Create a set of models with different degrees or any other variants.</li>
<li>Iterate through the $\lambda$s and for each $\lambda$ go through all the models to learn some $\Theta$.</li>
<li>Compute the cross validation error using the learned Θ (computed with λ) on the $J_{CV}(\Theta)$  <strong>without</strong>  regularization or λ = 0.</li>
<li>Select the best combo that produces the lowest error on the cross validation set.</li>
<li>Using the best combo Θ and λ, apply it on $J_{test}(\Theta)$ to see if it has a good generalization of the problem.<br>简单说：<br>训练12个不同归一化的模型-&gt;分别对应交叉验证集计算误差-&gt;选出最小的那个-&gt;使用在测试集上<h2 id="Regularization-相关结论"><a href="#Regularization-相关结论" class="headerlink" title="Regularization 相关结论"></a>Regularization 相关结论</h2><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170511/111820389.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure></li>
<li>当$\lambda$较小时，训练集误差较小（过拟合）而交叉验证集误差较大。</li>
<li>随着$\lambda$增加，训练集误差不断增加（欠拟合），而交叉验证集误差则是先减小后增大。</li>
</ul>
<h1 id="Learning-Curves"><a href="#Learning-Curves" class="headerlink" title="Learning Curves"></a>Learning Curves</h1><ul>
<li>学习曲线是一个很好的工具，我们会经常使用学习曲线来判断某一个学习算法是否处于偏差、方差问题。</li>
<li>学习曲线试讲训练集误差和交叉验证集误差作为训练集<strong>实例数量（m）</strong>的函数绘制的图表。</li>
</ul>
<h2 id="Experiencing-high-bias"><a href="#Experiencing-high-bias" class="headerlink" title="Experiencing high bias:"></a>Experiencing high bias:</h2><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170511/113614575.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<ul>
<li> **Low training set size**: causes $J_{train}(\Theta)$ to be low and $J_{CV}(\Theta)$ to be high. </li>
<li> **Large training set size**: causes both $J_{train}(\Theta)$ and $J_{CV}(\Theta)$ to be high with $J_{train}(\Theta)$≈$J_{CV}(\Theta)$ 
因此在高偏差（<strong>欠拟合</strong>）的情况下，增加训练集数量并不是一个好办法。此时，我们应当增加features。<h2 id="Experiencing-high-variance"><a href="#Experiencing-high-variance" class="headerlink" title="Experiencing high variance:"></a>Experiencing high variance:</h2><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170511/113828788.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure></li>
<li> **Low training set size**: $J_{train}(\Theta)$ will be low and $J_{CV}(\Theta)$ will be high. </li>
<li> **Large training set size**: $J_{train}(\Theta)$ increases with training set size and $J_{CV}(\Theta)$ continues to decrease without leveling off. Also, $J_{train}(\Theta)$ &lt; $J_{CV}(\Theta)$ but the difference between them remains significant. 
对比之下，如果在高方差（<strong>过拟合</strong>）的情况下，增加训练集数量可以明显降低误差，提高算法效果。</li>
</ul>
<h1 id="决定下一步做什么"><a href="#决定下一步做什么" class="headerlink" title="决定下一步做什么"></a>决定下一步做什么</h1><ul>
<li>获得更多的训练实例——解决高方差<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170511/103200427.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="high biais and high variance" title="">
                </div>
                <div class="image-caption">high biais and high variance</div>
            </figure></li>
<li>尝试减少特征的数量——解决高方差</li>
<li>尝试获得更多的特征——解决高偏差</li>
<li>尝试增加多项式特征——解决高偏差<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170511/111820389.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure> </li>
<li>尝试减少归一化程度 λ—&gt;提高拟合准确度—&gt;解决高偏差</li>
<li>尝试增加归一化程度 λ—&gt;防止过拟合—&gt;解决高方差</li>
</ul>
]]></content>
    
    <summary type="html">
    
      &lt;blockquote&gt;
&lt;p&gt;假设你在开发一个机器学习系统，或者在改进一个机器学习系统的性能，应如何做？&lt;/p&gt;
&lt;/blockquote&gt;
&lt;hr&gt;
    
    </summary>
    
      <category term="Coursera ML" scheme="http://feiyang.li/categories/Coursera-ML/"/>
    
    
      <category term="Machine Learning" scheme="http://feiyang.li/tags/Machine-Learning/"/>
    
      <category term="读书笔记" scheme="http://feiyang.li/tags/%E8%AF%BB%E4%B9%A6%E7%AC%94%E8%AE%B0/"/>
    
  </entry>
  
  <entry>
    <title>Coursera ML(9)-Programming Exercise 4</title>
    <link href="http://feiyang.li/2017/05/10/Coursera-ML-9-Programming-Exercise-4/index.html"/>
    <id>http://feiyang.li/2017/05/10/Coursera-ML-9-Programming-Exercise-4/index.html</id>
    <published>2017-05-10T10:01:38.000Z</published>
    <updated>2017-05-11T01:21:35.719Z</updated>
    
    <content type="html"><![CDATA[<script src="/assets/js/DPlayer.min.js"> </script><script src="/assets/js/APlayer.min.js"> </script><blockquote>
<p>对应反向传播算法内容</p>
</blockquote>
<hr>
<a id="more"></a>
<h1 id="Programming-Exercise-4-Neural-Networks-Learning"><a href="#Programming-Exercise-4-Neural-Networks-Learning" class="headerlink" title="Programming Exercise 4 - Neural Networks Learning"></a>Programming Exercise 4 - Neural Networks Learning</h1><figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div><div class="line">10</div><div class="line">11</div><div class="line">12</div><div class="line">13</div><div class="line">14</div><div class="line">15</div><div class="line">16</div><div class="line">17</div><div class="line">18</div><div class="line">19</div><div class="line">20</div></pre></td><td class="code"><pre><div class="line"><span class="comment"># %load ../../../standard_import.txt</span></div><div class="line"><span class="keyword">import</span> pandas <span class="keyword">as</span> pd</div><div class="line"><span class="keyword">import</span> numpy <span class="keyword">as</span> np</div><div class="line"><span class="keyword">import</span> matplotlib <span class="keyword">as</span> mpl</div><div class="line"><span class="keyword">import</span> matplotlib.pyplot <span class="keyword">as</span> plt</div><div class="line"></div><div class="line"><span class="comment"># load MATLAB files</span></div><div class="line"><span class="keyword">from</span> scipy.io <span class="keyword">import</span> loadmat</div><div class="line"></div><div class="line">pd.set_option(<span class="string">'display.notebook_repr_html'</span>, <span class="keyword">False</span>)</div><div class="line">pd.set_option(<span class="string">'display.max_columns'</span>, <span class="keyword">None</span>)</div><div class="line">pd.set_option(<span class="string">'display.max_rows'</span>, <span class="number">150</span>)</div><div class="line">pd.set_option(<span class="string">'display.max_seq_items'</span>, <span class="keyword">None</span>)</div><div class="line"> </div><div class="line"><span class="comment">#%config InlineBackend.figure_formats = &#123;'pdf',&#125;</span></div><div class="line">%matplotlib inline</div><div class="line"></div><div class="line"><span class="keyword">import</span> seaborn <span class="keyword">as</span> sns</div><div class="line">sns.set_context(<span class="string">'notebook'</span>)</div><div class="line">sns.set_style(<span class="string">'darkgrid'</span>)</div></pre></td></tr></table></figure>
<h2 id="Load-MATLAB-datafiles"><a href="#Load-MATLAB-datafiles" class="headerlink" title="Load MATLAB datafiles"></a>Load MATLAB datafiles</h2><figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div></pre></td><td class="code"><pre><div class="line">data = loadmat(<span class="string">'data/ex4data1.mat'</span>)</div><div class="line">data.keys()</div></pre></td></tr></table></figure>
<pre><code>dict_keys([&apos;__header__&apos;, &apos;__version__&apos;, &apos;__globals__&apos;, &apos;X&apos;, &apos;y&apos;])
</code></pre><figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div></pre></td><td class="code"><pre><div class="line">y = data[<span class="string">'y'</span>]</div><div class="line"><span class="comment"># Add intercept</span></div><div class="line">X = np.c_[np.ones((data[<span class="string">'X'</span>].shape[<span class="number">0</span>],<span class="number">1</span>)), data[<span class="string">'X'</span>]]</div><div class="line"></div><div class="line">print(<span class="string">'X:'</span>,X.shape, <span class="string">'(with intercept)'</span>)</div><div class="line">print(<span class="string">'y:'</span>,y.shape)</div></pre></td></tr></table></figure>
<pre><code>X: (5000, 401) (with intercept)
y: (5000, 1)
</code></pre><figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div></pre></td><td class="code"><pre><div class="line">weights = loadmat(<span class="string">'data/ex3weights.mat'</span>)</div><div class="line">weights.keys()</div></pre></td></tr></table></figure>
<pre><code>dict_keys([&apos;__header__&apos;, &apos;__version__&apos;, &apos;__globals__&apos;, &apos;Theta1&apos;, &apos;Theta2&apos;])
</code></pre><figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div></pre></td><td class="code"><pre><div class="line">theta1, theta2 = weights[<span class="string">'Theta1'</span>], weights[<span class="string">'Theta2'</span>]</div><div class="line">print(<span class="string">'theta1 :'</span>, theta1.shape)</div><div class="line">print(<span class="string">'theta2 :'</span>, theta2.shape)</div><div class="line">params = np.r_[theta1.ravel(), theta2.ravel()]</div><div class="line">print(<span class="string">'params :'</span>, params.shape)</div></pre></td></tr></table></figure>
<pre><code>theta1 : (25, 401)
theta2 : (10, 26)
params : (10285,)
</code></pre><h2 id="Neural-Network"><a href="#Neural-Network" class="headerlink" title="Neural Network"></a>Neural Network</h2><p>Input layer size = 400 (20x20 pixels) <br><br>Hidden layer size = 25 <br><br>Number of labels = 10</p>
<h1 id="Neural-Networks-Feed-Forward-and-Cost-Function"><a href="#Neural-Networks-Feed-Forward-and-Cost-Function" class="headerlink" title="Neural Networks - Feed Forward and Cost Function"></a>Neural Networks - Feed Forward and Cost Function</h1><figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div></pre></td><td class="code"><pre><div class="line"><span class="function"><span class="keyword">def</span> <span class="title">sigmoid</span><span class="params">(z)</span>:</span></div><div class="line">    <span class="keyword">return</span>(<span class="number">1</span> / (<span class="number">1</span> + np.exp(-z)))</div></pre></td></tr></table></figure>
<h2 id="Sigmoid-gradient"><a href="#Sigmoid-gradient" class="headerlink" title="Sigmoid gradient"></a>Sigmoid gradient</h2> $$ g'(z) = g(z)(1 - g(z))$$ 
<p>where  $$ g(z) = \frac{1}{1+e^{-z}}$$ </p>
<figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div></pre></td><td class="code"><pre><div class="line"><span class="function"><span class="keyword">def</span> <span class="title">sigmoidGradient</span><span class="params">(z)</span>:</span></div><div class="line">    <span class="keyword">return</span>(sigmoid(z)*(<span class="number">1</span>-sigmoid(z)))</div></pre></td></tr></table></figure>
<h2 id="Cost-Function"><a href="#Cost-Function" class="headerlink" title="Cost Function"></a>Cost Function</h2> $$ J(\theta) = \frac{1}{m}\sum_{i=1}^{m}\sum_{k=1}^{K}\big[-y^{(i)}_{k}\, log\,(( h_\theta\,(x^{(i)}))_k)-(1-y^{(i)}_k)\,log\,(1-h_\theta(x^{(i)}))_k)\big]$$ 
<h2 id="Regularized-Cost-Function"><a href="#Regularized-Cost-Function" class="headerlink" title="Regularized Cost Function"></a>Regularized Cost Function</h2> $$ J(\theta) = \frac{1}{m}\sum_{i=1}^{m}\sum_{k=1}^{K}\bigg[-y^{(i)}_{k}\, log\,(( h_\theta\,(x^{(i)}))_k)-(1-y^{(i)}_k)\,log\,(1-h_\theta(x^{(i)}))_k)\bigg] + \frac{\lambda}{2m}\bigg[\sum_{j=1}^{25}\sum_{k=1}^{400}(\Theta_{j,k}^{(1)})^2+\sum_{j=1}^{10}\sum_{k=1}^{25}(\Theta_{j,k}^{(2)})^2\bigg]$$ 
<figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div><div class="line">10</div><div class="line">11</div><div class="line">12</div><div class="line">13</div><div class="line">14</div><div class="line">15</div><div class="line">16</div><div class="line">17</div><div class="line">18</div><div class="line">19</div><div class="line">20</div><div class="line">21</div><div class="line">22</div><div class="line">23</div><div class="line">24</div><div class="line">25</div><div class="line">26</div><div class="line">27</div><div class="line">28</div><div class="line">29</div><div class="line">30</div><div class="line">31</div><div class="line">32</div><div class="line">33</div><div class="line">34</div><div class="line">35</div><div class="line">36</div></pre></td><td class="code"><pre><div class="line"><span class="function"><span class="keyword">def</span> <span class="title">nnCostFunction</span><span class="params">(nn_params, input_layer_size, hidden_layer_size, num_labels, features, classes, reg)</span>:</span></div><div class="line">    </div><div class="line">    <span class="comment"># When comparing to Octave code note that Python uses zero-indexed arrays.</span></div><div class="line">    <span class="comment"># But because Numpy indexing does not include the right side, the code is the same anyway.</span></div><div class="line">    theta1 = nn_params[<span class="number">0</span>:(hidden_layer_size*(input_layer_size+<span class="number">1</span>))].reshape(hidden_layer_size,(input_layer_size+<span class="number">1</span>))</div><div class="line">    theta2 = nn_params[(hidden_layer_size*(input_layer_size+<span class="number">1</span>)):].reshape(num_labels,(hidden_layer_size+<span class="number">1</span>))</div><div class="line"></div><div class="line">    m = features.shape[<span class="number">0</span>]</div><div class="line">    y_matrix = pd.get_dummies(classes.ravel()).as_matrix() </div><div class="line">    </div><div class="line">    <span class="comment"># Cost</span></div><div class="line">    a1 = features <span class="comment"># 5000x401</span></div><div class="line">        </div><div class="line">    z2 = theta1.dot(a1.T) <span class="comment"># 25x401 * 401x5000 = 25x5000 </span></div><div class="line">    a2 = np.c_[np.ones((features.shape[<span class="number">0</span>],<span class="number">1</span>)),sigmoid(z2.T)] <span class="comment"># 5000x26 </span></div><div class="line">    </div><div class="line">    z3 = theta2.dot(a2.T) <span class="comment"># 10x26 * 26x5000 = 10x5000 </span></div><div class="line">    a3 = sigmoid(z3) <span class="comment"># 10x5000</span></div><div class="line">    </div><div class="line">    J = <span class="number">-1</span>*(<span class="number">1</span>/m)*np.sum((np.log(a3.T)*(y_matrix)+np.log(<span class="number">1</span>-a3).T*(<span class="number">1</span>-y_matrix))) + \</div><div class="line">        (reg/(<span class="number">2</span>*m))*(np.sum(np.square(theta1[:,<span class="number">1</span>:])) + np.sum(np.square(theta2[:,<span class="number">1</span>:])))</div><div class="line"></div><div class="line">    <span class="comment"># Gradients</span></div><div class="line">    d3 = a3.T - y_matrix <span class="comment"># 5000x10</span></div><div class="line">    d2 = theta2[:,<span class="number">1</span>:].T.dot(d3.T)*sigmoidGradient(z2) <span class="comment"># 25x10 *10x5000 * 25x5000 = 25x5000</span></div><div class="line">    </div><div class="line">    delta1 = d2.dot(a1) <span class="comment"># 25x5000 * 5000x401 = 25x401</span></div><div class="line">    delta2 = d3.T.dot(a2) <span class="comment"># 10x5000 *5000x26 = 10x26</span></div><div class="line">    </div><div class="line">    theta1_ = np.c_[np.ones((theta1.shape[<span class="number">0</span>],<span class="number">1</span>)),theta1[:,<span class="number">1</span>:]]</div><div class="line">    theta2_ = np.c_[np.ones((theta2.shape[<span class="number">0</span>],<span class="number">1</span>)),theta2[:,<span class="number">1</span>:]]</div><div class="line">    </div><div class="line">    theta1_grad = delta1/m + (theta1_*reg)/m</div><div class="line">    theta2_grad = delta2/m + (theta2_*reg)/m</div><div class="line">    </div><div class="line">    <span class="keyword">return</span>(J, theta1_grad, theta2_grad)</div></pre></td></tr></table></figure>
<figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div></pre></td><td class="code"><pre><div class="line"><span class="comment"># Regularization parameter = 0</span></div><div class="line">nnCostFunction(params, <span class="number">400</span>, <span class="number">25</span>, <span class="number">10</span>, X, y, <span class="number">0</span>)[<span class="number">0</span>]</div></pre></td></tr></table></figure>
<pre><code>0.28762916516131892
</code></pre><figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div></pre></td><td class="code"><pre><div class="line"><span class="comment"># Regularization parameter = 1</span></div><div class="line">nnCostFunction(params, <span class="number">400</span>, <span class="number">25</span>, <span class="number">10</span>, X, y, <span class="number">1</span>)[<span class="number">0</span>]</div></pre></td></tr></table></figure>
<pre><code>0.38376985909092365
</code></pre><figure class="highlight python"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">[sigmoidGradient(z) <span class="keyword">for</span> z <span class="keyword">in</span> [<span class="number">-1</span>, <span class="number">-0.5</span>, <span class="number">0</span>, <span class="number">0.5</span>, <span class="number">1</span>]]</div></pre></td></tr></table></figure>
<pre><code>[0.19661193324148185,
 0.23500371220159449,
 0.25,
 0.23500371220159449,
 0.19661193324148185]
</code></pre>]]></content>
    
    <summary type="html">
    
      &lt;blockquote&gt;
&lt;p&gt;对应反向传播算法内容&lt;/p&gt;
&lt;/blockquote&gt;
&lt;hr&gt;
    
    </summary>
    
      <category term="Coursera ML" scheme="http://feiyang.li/categories/Coursera-ML/"/>
    
    
      <category term="Machine Learning" scheme="http://feiyang.li/tags/Machine-Learning/"/>
    
      <category term="读书笔记" scheme="http://feiyang.li/tags/%E8%AF%BB%E4%B9%A6%E7%AC%94%E8%AE%B0/"/>
    
      <category term="Python" scheme="http://feiyang.li/tags/Python/"/>
    
  </entry>
  
  <entry>
    <title>ss-panel 一键安装脚本（前端后端全自动一键安装脚本）</title>
    <link href="http://feiyang.li/2017/05/09/ss-shell/index.html"/>
    <id>http://feiyang.li/2017/05/09/ss-shell/index.html</id>
    <published>2017-05-09T13:58:57.000Z</published>
    <updated>2017-06-03T02:14:30.661Z</updated>
    
    <content type="html"><![CDATA[<script src="/assets/js/DPlayer.min.js"> </script><script src="/assets/js/APlayer.min.js"> </script><blockquote>
<p>上周接触了ss-panel，发现搭建起来异常的麻烦，对新手来说，到处都是坑，因此写了本ss-panel一键安装脚本 。示例站点：<a href="https://ss.91vps.club/" target="_blank" rel="external">https://ss.91vps.club/</a></p>
</blockquote>
<hr>
<a id="more"></a>
<p>你们要的ss-panel v3 mod 脚本已经写好了，<a href="https://91vps.club/2017/05/27/ss-panel-v3-mod/" target="_blank" rel="external">点击查看</a></p>
<h1 id="特点"><a href="#特点" class="headerlink" title="特点"></a>特点</h1><ul>
<li>可以使用BBR加速：选择对应<a href="http://feiyang.li/2017/05/05/ss-panel-full/index.html#OpenVZ-以外的（-KVM-、-Xen-、-VMware-等）">一键代码</a>即可</li>
<li>使用<code>supervisord</code>，守护运行。（重启后输入<code>supervisord</code>，即可启动守护程序）</li>
<li>前端使用最新<a href="https://github.com/orvice/ss-panel" target="_blank" rel="external">ss-panel v3</a>，稳定性和可管理行都有明显提高。</li>
<li>后端使用<a href="https://github.com/fsgmhoward/shadowsocks-py-mu" target="_blank" rel="external">shadowsocks-py-mu</a>，多用户版本，与前端完美对接。</li>
<li>过程全自动，所以可能碰到的坑都提前做好处理。</li>
<li>被墙掉的资源都换成了墙内资源，不会出现被卡死的情况了。</li>
</ul>
<hr>
<h1 id="效果"><a href="#效果" class="headerlink" title="效果"></a>效果</h1><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170509/191015542.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170509/191042466.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170509/191103228.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<p><strong>Github：<a href="https://github.com/mmmwhy/ss-panel-and-ss-py-mu" target="_blank" rel="external">https://github.com/mmmwhy/ss-panel-and-ss-py-mu</a></strong></p>
<p><strong>示例网站：<a href="https://ss.feiyang.li/" target="_blank" rel="external">https://ss.feiyang.li/</a></strong></p>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170509/215724204.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<h1 id="系统要求"><a href="#系统要求" class="headerlink" title="系统要求"></a>系统要求</h1><p>建议CentOS7 X64，我用的是这个版本，目前在腾讯云，digitalocean，interserver上通过测试。<br>理论上：CentOS 6+ / Debian 7+ / Ubuntu 14.04 +  都可以，</p>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170510/094150095.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<h1 id="一键安装脚本"><a href="#一键安装脚本" class="headerlink" title="一键安装脚本"></a>一键安装脚本</h1><ul>
<li>登陆后运行：<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">screen -S ss</div></pre></td></tr></table></figure>
</li>
</ul>
<p>如果提示screen: command not found 命令不存在可以执行：<code>yum install screen</code> 或 <code>apt-get install screen</code>安装（如果网络掉线，可以重新连接，再执行 <code>screen -r ss</code> 就会看到你的ss-panel安装进程。）</p>
<ul>
<li>安装脚本<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">wget -N --no-check-certificate https://raw.githubusercontent.com/mmmwhy/ss-panel-and-ss-py-mu/master/ss-panel_node.sh &amp;&amp; chmod +x ss-panel_node.sh &amp;&amp; bash ss-panel_node.sh</div></pre></td></tr></table></figure>
</li>
</ul>
<p>运行脚本后会出现脚本操作菜单，</p>
<h2 id="ss-panel和ss-node同时安装"><a href="#ss-panel和ss-node同时安装" class="headerlink" title="ss-panel和ss-node同时安装"></a>ss-panel和ss-node同时安装</h2><p>选择并输入 1 </p>
<p><img src="http://cdn.mmmxcc.cn/blog/20170509/214909086.png" alt=""></p>
<p>因为ss-panel依赖于mysql,php,nginx，使用lnmp一键安装包，点击任意键开始安装。<br><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170510/102436162.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure></p>
<p>LNMP包编译时间较长，可以喝杯茶吃个饭。大约30分钟左右，安装结束，提示登陆IP即可查看网站</p>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170510/102100972.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170510/101919599.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<p><strong>默认账户：ss@feiyang.li</strong></p>
<p><strong>默认密码：feiyang</strong></p>
<p>进入ss-panel页面后，记得在<strong>管理面板-&gt;节点管理-&gt;添加节点-&gt;输入节点信息</strong></p>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170510/085511290.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<p>之后回到用户面板，就可以使用了。</p>
<h2 id="仅安装ss-panel"><a href="#仅安装ss-panel" class="headerlink" title="仅安装ss-panel"></a>仅安装ss-panel</h2><p>选择并输入 2<br>出现的结果与1相同，只不过您vps上没有同时安装<a href="https://github.com/fsgmhoward/shadowsocks-py-mu" target="_blank" rel="external">ss-py-mu</a></p>
<h2 id="仅安装ss-node"><a href="#仅安装ss-node" class="headerlink" title="仅安装ss-node"></a>仅安装ss-node</h2><ul>
<li>先在网页增加节点信息，特别要记住这里的node_id，长这个样子的</li>
</ul>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170509/221038086.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<ul>
<li>选择并输入 3，用于新建节点。</li>
</ul>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170509/221216262.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<ul>
<li>输入相关信息，ip地址和域名（ss-panel的）都可以，但是需要加上<strong>http:// 或者 https://</strong> ，注意区分自己域名有没有ssl。否则可能出现用户使用记录无法推送的问题。</li>
<li>如果使用本一键脚本搭建前端页面，第二个mukey可以直接按回车，默认是<strong>mupass</strong></li>
<li>第三个位置输入之前那个node_id。</li>
<li>如果最后提示cat log失败，其实是不影响的，因为有没有记录是不影响的。<h2 id="查看-shadowsocks-manyuser-是否已经运行："><a href="#查看-shadowsocks-manyuser-是否已经运行：" class="headerlink" title="查看 shadowsocks-manyuser 是否已经运行："></a>查看 shadowsocks-manyuser 是否已经运行：</h2><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div></pre></td><td class="code"><pre><div class="line">ps -ef | grep servers.py</div><div class="line">root 952 739 0 15:40 ? 00:00:00 python /root/shadowsocks-rm/shadowsocks/servers.p</div></pre></td></tr></table></figure>
</li>
</ul>
<p>出现底下那一行，就代表运行成功了。</p>
<h1 id="其他补充内容"><a href="#其他补充内容" class="headerlink" title="其他补充内容"></a>其他补充内容</h1><ul>
<li>LNMP环境编译时间较长</li>
<li>php依赖安装会花费较长时间，特别是在国内<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170510/101054745.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="本页面停留时间较长" title="">
                </div>
                <div class="image-caption">本页面停留时间较长</div>
            </figure></li>
<li>如果想设置更多的信息，请查看<a href="http://feiyang.li/2017/05/05/ss-panel-full/index.html">安装教程全文</a>，本脚本和该教程内容完全一致。</li>
<li><a href="http://feiyang.li/2017/05/03/ss-panel/index.html#常见错误">常见错误</a>在这里查看</li>
<li>不停的有人问“node不能添加”“启动不起来”“怎么添加域名”“怎么修改签到流量”“怎么修改初始流量”…….如果真的有什么问题，建议先看上边两个页面，其实写的很清楚了。</li>
<li>如果总是遇到不能连接的问题，建议直接<a href="http://feiyang.li/2017/05/03/ss-panel/index.html#关闭防火墙">关闭node节点的防火墙</a></li>
</ul>
]]></content>
    
    <summary type="html">
    
      &lt;blockquote&gt;
&lt;p&gt;上周接触了ss-panel，发现搭建起来异常的麻烦，对新手来说，到处都是坑，因此写了本ss-panel一键安装脚本 。示例站点：&lt;a href=&quot;https://ss.91vps.club/&quot;&gt;https://ss.91vps.club/&lt;/a&gt;&lt;/p&gt;
&lt;/blockquote&gt;
&lt;hr&gt;
    
    </summary>
    
      <category term="shadowsocks" scheme="http://feiyang.li/categories/shadowsocks/"/>
    
    
      <category term="GFW" scheme="http://feiyang.li/tags/GFW/"/>
    
      <category term="Linux" scheme="http://feiyang.li/tags/Linux/"/>
    
      <category term="教程" scheme="http://feiyang.li/tags/%E6%95%99%E7%A8%8B/"/>
    
      <category term="vps" scheme="http://feiyang.li/tags/vps/"/>
    
      <category term="shell" scheme="http://feiyang.li/tags/shell/"/>
    
  </entry>
  
  <entry>
    <title>Coursera ML(8)-Neural Networks: Learning</title>
    <link href="http://feiyang.li/2017/05/06/Coursera-ML-8-Neural-Networks-Learning/index.html"/>
    <id>http://feiyang.li/2017/05/06/Coursera-ML-8-Neural-Networks-Learning/index.html</id>
    <published>2017-05-06T13:52:23.000Z</published>
    <updated>2017-05-10T11:56:05.596Z</updated>
    
    <content type="html"><![CDATA[<script src="/assets/js/DPlayer.min.js"> </script><script src="/assets/js/APlayer.min.js"> </script><blockquote>
<p>本节笔记对应第五周Coursera课程 <a href="https://www.coursera.org/learn/machine-learning/home/week/5" target="_blank" rel="external">Neural Networks: Learning</a></p>
</blockquote>
<hr>
<a id="more"></a>
<h1 id="Cost-Fuction"><a href="#Cost-Fuction" class="headerlink" title="Cost Fuction"></a>Cost Fuction</h1><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170504/221046501.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<ul>
<li>cost function  $$\begin{gather*} J(\Theta) = - \frac{1}{m} \sum_{i=1}^m \sum_{k=1}^K \left[y^{(i)}_k \log ((h_\Theta (x^{(i)}))_k) + (1 - y^{(i)}_k)\log (1 - (h_\Theta(x^{(i)}))_k)\right] \end{gather*}$$ 
我们使用循环，对每一行特征都预测K个不同的结果，然后利用循环在K个预测中选择可能性最高的一个，将其与y中的实际数据进行对比。</li>
</ul>
<ul>
<li>Similarly, the regularization term just sum up the square of all the parameters from all the layers: $$\frac{\lambda}{2m}\sum_{l=1}^{L-1} \sum_{i=1}^{s_l} \sum_{j=1}^{s_{l+1}} ( \Theta_{j,i}^{(l)})^2$$ 
正则化的这一项，是在排除了每一层$\theta_0$后，每一层的$\Theta$矩阵的和。</li>
<li>最里层的循环$j$循环所有的行(由$s_{l+1}$层的激活单元数决定)。</li>
<li>循环$i$则循环所有的列，由该层(s_l层)的激活单元数决定。</li>
</ul>
<p>Note:<br>这个看起来很复杂的代价函数与之前代价函数思想是一样的，我们希望通过代价函数来观察算法预测的结果与真实情况的误差有多大。不同的是，对于每一行特征，我们都会给出<strong>K个预测</strong>。</p>
<h1 id="Backpropagation-Algorithm"><a href="#Backpropagation-Algorithm" class="headerlink" title="Backpropagation Algorithm"></a>Backpropagation Algorithm</h1><p><del>这一块我也看的不是很懂，先不做笔记。之后如果有新的感悟，再记回来。</del></p>
<p><del>没有就算了。</del></p>
<p>最终证明还是可以看懂的，csdn这个sb编辑器，害得我记得笔记丢失了一次，好气哦。又得抄大佬的笔记了。</p>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170508/090514790.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<p>In order to use optimization algorithms like gradient descent, we must calculate the derivative of the cost function. </p>
<ul>
<li>First we should calculate $\delta$ for each layer using backpropagation algorithm.  From $layer_l$ to $layer_{l+1}$ 
we have: $$\delta^{(output)} = a^{(output)} - y\\\delta^{(l)} = [\Theta^{(l)}]^T \delta^{(l+1)} .* g'(z^{(l)})$$  
</li>
</ul>
<p>误差$\delta$表示激活单元的预测值与实际值之间的误差。第二行式子，是在后一层偏差的基础上，增加了本层的偏差。</p>
<p>注意：没有$\delta_1$</p>
<h2 id="一点感悟"><a href="#一点感悟" class="headerlink" title="一点感悟"></a>一点感悟</h2><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170508/104937137.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170508/104732466.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<p>将这两个图，联系在一起，好像对反向传播有了一些理解。其实就是在正向传播的基础上，因为最终的对比结果是在最外层，那么我们从最外层开始，重新纠正一遍计算过程，这样可以减少误差。</p>
<ul>
<li><p>For sigmoid function, $g’(z) = g(z)(1-g(z))$<br>这个比较简单，就不推导了。</p>
</li>
<li><p>Then we have the derivative:</p>
 $$\frac {\partial}{\partial \Theta_{ij}^{(l)}} J(\Theta) = a_j^{(l)} \delta_i^{(l+1)} \\\Delta^{(l)} = \delta^{(l+1)} (a^{(l)})^T$$ 
<p>在Coursera教程中，这个地方被草草带过，相比之下<a href="http://www.ics.uci.edu/~dramanan/teaching/ics273a_winter08/lectures/lecture14.pdf" target="_blank" rel="external">uci</a>的这篇文章介绍的非常详细。</p>
</li>
</ul>
<h2 id="一个推导过程"><a href="#一个推导过程" class="headerlink" title="一个推导过程"></a>一个推导过程</h2><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170508/142800870.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170508/142813675.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170508/142827146.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170508/142832567.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<h2 id="另外一个推导"><a href="#另外一个推导" class="headerlink" title="另外一个推导"></a>另外一个推导</h2><h3 id="BackPropagation算法"><a href="#BackPropagation算法" class="headerlink" title="BackPropagation算法"></a>BackPropagation算法</h3><p>BackPropagation算法是多层神经网络的训练中举足轻重的算法，简单的理解，它就是复合函数的链式法则。由于后面我的网络中会用到对数损失函数，所以在这里我们使用平方损失函数。对于单个样例，其平方损失函数为：<br><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170508/144354206.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure></p>
<p>对于给定一个包含m个样例的数据集，我们可以定义整体代价函数为：<br><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170508/144400651.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure><br>和直线的拟合类似，深度学习也有一个目标函数，通过这个目标函数我们可以知道参数为何值时对我们来说才是一组“好”的参数，这个函数就是前边提到的损失函数。训练的过程就是通过每一次迭代对网络中参数进行更新，来使损失函数的值达到最小(下图中α为学习率)。<br><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170508/144431808.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure></p>
<h3 id="利用BackPropagation算法计算偏导数"><a href="#利用BackPropagation算法计算偏导数" class="headerlink" title="利用BackPropagation算法计算偏导数"></a>利用BackPropagation算法计算偏导数</h3><p>由上一节可知，我们只需求出每一层的 <figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170508/144537773.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="](http://cdn.mmmxcc.cn/blog/20170508/144525573.png) 和![" title="">
                </div>
                <div class="image-caption">](http://cdn.mmmxcc.cn/blog/20170508/144525573.png) 和![</div>
            </figure> 即可完成该层的权值和偏置的更新。<br>BP算法的整体思路如下：对于每个给定的训练样本，首先进行前向计算，计算出网络中每一层的激活值和网络的输出。对于最后一层(输出层)，我们可以直接计算出网络的输出值与已经给出的标签值(label)直接的差距，我们将这个值定义为残差δ。对于输出层之前的隐藏层L，我们将根据L+1层各节点的加权平均值来计算第L层的残差。</p>
<p><strong>插入一些我个人对BP算法的一点比较容易理解的解释(如有错误请指出)：在反向传播过程中，若第x层的a节点通过权值W对x+1层的b节点有贡献，则在反向传播过程中，梯度通过权值W从b节点传播回a节点。不管下面的公式推导，还是后面的卷积神经网络，在反向传播的过程中，都是遵循这样的一个规律。</strong></p>
<p>反向传播的具体步骤如下：</p>
<ul>
<li>根据输入，计算出每一层的激活值。</li>
<li>对于输出层，我们使用如下公式计算每个单元的残差：<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170508/145300526.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure></li>
<li>对于输出层之前的每一层，按照如下公式计算每一层的残差：<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170508/145319283.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure></li>
<li>由残差计算每一层的偏导数：<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170508/145334453.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure></li>
<li>最后，使用偏导数更新权值和偏置。</li>
</ul>
<hr>
<p>其实说到底，这两个推导方法其实是一回事，只不过换了几个字母而已。第二个推导更好一些</p>
<ul>
<li>Instead of calculate the $\Delta$  for all the training data, we could sum up all the inputs independantly, like:<br>From i = 1 to m, do:<br>$$\Delta^{(l)} += \delta^{(l+1)} (a^{(l)})^T $$<br>Then we have: $$ \\\frac {\partial}{\partial \Theta_{ij}^{(l)}} J(\Theta) = \frac {1}{m} \Delta_{ij}^{(l)}$$ 
</li>
</ul>
<p>(This step is the main step for mapreduce, see Large Scale Machine Learning)</p>
<ul>
<li>Then add regularzation term for partial derivative, we have:</li>
</ul>
<p>$ $\frac {\partial}{\partial \Theta_{ij}^{(l)}} J(\Theta) = \frac {1}{m} \Delta_{ij}^{(l)} + \frac {\lambda}{m} \Theta_{ij}^{(l)}$$ </p>
<ul>
<li>Finally, using $J(\Theta)$ and $\frac {\partial}{\partial \Theta_{ij}^{(l)}} J(\Theta)$, we could minimize the cost function using gradient descent as well as other optimization algorithms.</li>
</ul>
<h1 id="Optional-Section-How-Backpropagation-Works"><a href="#Optional-Section-How-Backpropagation-Works" class="headerlink" title="Optional Section: How Backpropagation Works"></a>Optional Section: How Backpropagation Works</h1><p>Using the same model as the last section:<br> $$\begin{align}\frac {\partial J}{\partial \Theta_{ij}^{(1)}} =\frac {\partial J}{\partial a_i^{(2)}}\frac {\partial a_i^{(2)}}{\partial z_i^{(2)}}\frac {\partial z_i^{(2)}}{\partial \Theta_{ij}^{(1)}} \\ =\frac {\partial J}{\partial a_i^{(2)}} g’(z_i^{(2)}) a_i^{(1)}\end{align}$$Then taking the total derivative with respect to $z^{(3)}$, a recursive expression for the derivative is obtained:$$\begin{align}\frac {\partial J}{\partial a_i^{(2)}} =\sum\limits_{m=1}^{s_3}\frac {\partial J}{\partial z_m^{(3)}}\frac {\partial z_m^{(3)}}{\partial a_i^{(2)}} \\ =\sum\limits_{m=1}^{s_3}\frac {\partial J}{\partial a_m^{(3)}}\frac {\partial a_m^{(3)}}{\partial z_m^{(3)}}\frac {\partial z_m^{(3)}}{\partial a_i^{(2)}} \\ =\sum\limits_{m=1}^{s_3}\frac {\partial J}{\partial a_m^{(3)}} g’(z_m^{(3)}) \Theta_{mi}^{(2)}\end{align}$$<br>Let  $\delta_i^{(2)} =\frac {\partial J}{\partial a_i^{2}} g’(z_i^{(2)})$, then from the result of the first formula We have: $\frac {\partial J}{\partial \Theta_{ij}^{1}} =\delta_i^{(2)} a_i^{(1)}$, in which recursively, $\delta_i^{(2)} =(\sum\limits_{m=1}^{s_3}\delta_m^{(3)}\Theta_{mi}^{(2)}) g’(z_i^{(2)})$ </p>
<p>–</p>
<p>Referance:<br><a href="http://www.wikiwand.com/en/Backpropagation#Derivation" target="_blank" rel="external">http://www.wikiwand.com/en/Backpropagation#Derivation</a><br>ttp://blog.csdn.net/l691899397/article/details/52223998<br><a href="http://xxuan.me/2016-02-20-Neural-Networks-Learning.html" target="_blank" rel="external">http://xxuan.me/2016-02-20-Neural-Networks-Learning.html</a></p>
]]></content>
    
    <summary type="html">
    
      &lt;blockquote&gt;
&lt;p&gt;本节笔记对应第五周Coursera课程 &lt;a href=&quot;https://www.coursera.org/learn/machine-learning/home/week/5&quot;&gt;Neural Networks: Learning&lt;/a&gt;&lt;/p&gt;
&lt;/blockquote&gt;
&lt;hr&gt;
    
    </summary>
    
      <category term="Coursera ML" scheme="http://feiyang.li/categories/Coursera-ML/"/>
    
    
      <category term="Machine Learning" scheme="http://feiyang.li/tags/Machine-Learning/"/>
    
      <category term="读书笔记" scheme="http://feiyang.li/tags/%E8%AF%BB%E4%B9%A6%E7%AC%94%E8%AE%B0/"/>
    
      <category term="Python" scheme="http://feiyang.li/tags/Python/"/>
    
  </entry>
  
  <entry>
    <title>shadowsocks-Clients</title>
    <link href="http://feiyang.li/2017/05/05/ss-client/index.html"/>
    <id>http://feiyang.li/2017/05/05/ss-client/index.html</id>
    <published>2017-05-05T06:23:13.000Z</published>
    <updated>2017-05-20T01:22:29.813Z</updated>
    
    <content type="html"><![CDATA[<script src="/assets/js/DPlayer.min.js"> </script><script src="/assets/js/APlayer.min.js"> </script><blockquote>
<p> shadowsocks是目前fq速度最快的工具之一，可以转发udp。可以使用ipv6，对于很多高校学生来说，可以通过此方法节省流量费用。</p>
</blockquote>
<hr>
<a id="more"></a>
<h1 id="Windows"><a href="#Windows" class="headerlink" title="Windows"></a>Windows</h1><h2 id="ss"><a href="#ss" class="headerlink" title="ss"></a>ss</h2><p><a href="http://home.ustc.edu.cn/~mmmwhy/Shadowsocks.exe" target="_blank" rel="external">http://home.ustc.edu.cn/~mmmwhy/Shadowsocks.exe</a></p>
<ul>
<li>在任务栏找到 Shadowsocks 图标</li>
<li>在 服务器 菜单添加多个服务器</li>
<li>选择 启用系统代理 来启用系统代理。请禁用浏览器里的代理插件，或把它们设置为使用系统代理。<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170515/104128278.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<h2 id="ssr"><a href="#ssr" class="headerlink" title="ssr"></a>ssr</h2><a href="http://home.ustc.edu.cn/~mmmwhy/ShadowsocksR-dotnet4.0.exe" target="_blank" rel="external">http://home.ustc.edu.cn/~mmmwhy/ShadowsocksR-dotnet4.0.exe</a><br>右键 纸飞机 -&gt; 扫描二维码 -&gt;  开启系统代理选择 -&gt;pac模式或者全局模式<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170515/104039212.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
</li>
</ul>
<p>如果遇到某个节点不能使用，可以换一个，因为本地运营商可能会把某些ip墙掉。</p>
<h1 id="Mac-OS-X"><a href="#Mac-OS-X" class="headerlink" title="Mac OS X"></a>Mac OS X</h1><p><a href="https://github.com/shadowsocks/ShadowsocksX-NG/releases" target="_blank" rel="external">https://github.com/shadowsocks/ShadowsocksX-NG/releases</a><br>OS X 下设置使用方法<br>Mac 下使用极其方便，因为 GUI 已经集成了系统 Pac，所以无需设置任何科学上网的姿势，直接运行 ShadowsocksX ，然后输入服务商提供给你的账号信息，保存运行后，即可在所有浏览器里直接科学上网，如图：<br><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170515/105527529.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure><br><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170515/105543198.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure></p>
<h1 id="Linux"><a href="#Linux" class="headerlink" title="Linux"></a>Linux</h1><p><a href="https://github.com/shadowsocks/shadowsocks-qt5/wiki/Installation" target="_blank" rel="external">https://github.com/shadowsocks/shadowsocks-qt5/wiki/Installation</a></p>
<h1 id="Android"><a href="#Android" class="headerlink" title="Android"></a>Android</h1><p><a href="http://cdn.mmmxcc.cn/ssr_3.0.2.2.apk" target="_blank" rel="external">ssr</a><br><figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170505/143305745.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure><br>扫描二维码-&gt; 然后点小飞机就可以了</p>
<h1 id="iOS"><a href="#iOS" class="headerlink" title="iOS"></a>iOS</h1><p><a href="https://itunes.apple.com/cn/app/ssrconnect-proxy-utility/id1161221960?mt=8" target="_blank" rel="external">SsrConnect - Proxy Utility</a> 免费<br><a href="https://itunes.apple.com/us/app/alice-network-proxy-utility/id1135320992?mt=8" target="_blank" rel="external">Network Proxy Utility</a> 免费<br><a href="https://itunes.apple.com/cn/app/shadowrocket/id932747118?mt=8" target="_blank" rel="external">Shadowrocket</a> 付费<br><a href="https://itunes.apple.com/app/apple-store/id1070901416?mt=8" target="_blank" rel="external">土豆丝</a> 付费</p>
<h1 id="OpenWRT"><a href="#OpenWRT" class="headerlink" title="OpenWRT"></a>OpenWRT</h1><ul>
<li>shadowsocks-libev<br><code>opkg install shadowsocks-libev</code></li>
<li>shadowsocks-libev-polarssl<br><code>opkg install shadowsocks-libev-polarssl</code></li>
</ul>
]]></content>
    
    <summary type="html">
    
      &lt;blockquote&gt;
&lt;p&gt; shadowsocks是目前fq速度最快的工具之一，可以转发udp。可以使用ipv6，对于很多高校学生来说，可以通过此方法节省流量费用。&lt;/p&gt;
&lt;/blockquote&gt;
&lt;hr&gt;
    
    </summary>
    
      <category term="shadowsocks" scheme="http://feiyang.li/categories/shadowsocks/"/>
    
    
      <category term="GFW" scheme="http://feiyang.li/tags/GFW/"/>
    
      <category term="Linux" scheme="http://feiyang.li/tags/Linux/"/>
    
      <category term="教程" scheme="http://feiyang.li/tags/%E6%95%99%E7%A8%8B/"/>
    
      <category term="vps" scheme="http://feiyang.li/tags/vps/"/>
    
  </entry>
  
  <entry>
    <title>ss-panel + ss py mu安装教程</title>
    <link href="http://feiyang.li/2017/05/05/ss-panel-full/index.html"/>
    <id>http://feiyang.li/2017/05/05/ss-panel-full/index.html</id>
    <published>2017-05-05T06:09:39.000Z</published>
    <updated>2017-05-26T01:13:00.278Z</updated>
    
    <content type="html"><![CDATA[<script src="/assets/js/DPlayer.min.js"> </script><script src="/assets/js/APlayer.min.js"> </script><blockquote>
<p>系统介绍： 1g内存，单核，也就是传说中的腾讯1元学生机。centos6.8 x32。</p>
</blockquote>
<hr>
<a id="more"></a>
<p>读者建议：如果使用国外vps<a href="http://feiyang.li/2017/05/03/ss-panel/index.html">建议</a>会快速很多。国内vps建议本方法。</p>
<h1 id="lnmp环境配置"><a href="#lnmp环境配置" class="headerlink" title="lnmp环境配置"></a>lnmp环境配置</h1><p>这里使用最新的<a href="https://lnmp.org/install.html" target="_blank" rel="external">lnmp1.4</a><br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">wget -c http://soft.vpser.net/lnmp/lnmp1.4beta.tar.gz &amp;&amp; tar zxf lnmp1.4beta.tar.gz &amp;&amp; cd lnmp1.4 &amp;&amp; ./install.sh lnmp</div></pre></td></tr></table></figure></p>
<p><img src="http://cdn.mmmxcc.cn/blog/20170504/204651581.png" alt=""></p>
<h1 id="安装ss-panel前端"><a href="#安装ss-panel前端" class="headerlink" title="安装ss-panel前端"></a>安装ss-panel前端</h1><h2 id="下载-ss-panel-源码"><a href="#下载-ss-panel-源码" class="headerlink" title="下载 ss-panel 源码"></a>下载 ss-panel 源码</h2><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div></pre></td><td class="code"><pre><div class="line">cd /home/wwwroot/ </div><div class="line">git clone https://github.com/orvice/ss-panel.git</div></pre></td></tr></table></figure>
<p>安装好之后，是这个样子的。</p>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170504/205330510.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="mark" title="">
                </div>
                <div class="image-caption">mark</div>
            </figure>
<h2 id="配置-ss-panel"><a href="#配置-ss-panel" class="headerlink" title="配置 ss-panel"></a>配置 ss-panel</h2><h3 id="env文件配置"><a href="#env文件配置" class="headerlink" title=".env文件配置"></a>.env文件配置</h3><p>安装依赖文件，并配置.env<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div></pre></td><td class="code"><pre><div class="line">cd /home/wwwroot/ss-panel/</div><div class="line">curl -sS https://install.phpcomposer.com/installer | php</div><div class="line">chmod +x composer.phar</div><div class="line">php composer.phar install</div></pre></td></tr></table></figure></p>
<ul>
<li>安装依赖那一步时间会久一些，因为一共有75个依赖，而且很多被墙了（因此如果直接用国外vps，真的会好很多），换了国内源会<strong>稍微</strong>好一些。<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170504/215216082.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="漫长的等待" title="">
                </div>
                <div class="image-caption">漫长的等待</div>
            </figure>
</li>
</ul>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div></pre></td><td class="code"><pre><div class="line">chmod -R 777 storage</div><div class="line">cp .env.example .env</div><div class="line">vi .env</div></pre></td></tr></table></figure>
<ul>
<li>这里主要是修改数据库的内容，还有<strong>muKey</strong>。</li>
</ul>
<h3 id="nginx配置"><a href="#nginx配置" class="headerlink" title="nginx配置"></a>nginx配置</h3><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div></pre></td><td class="code"><pre><div class="line">lnmp vhost add</div><div class="line"># 写入站点网址</div><div class="line">/home/wwwroot/ss-panel/public</div><div class="line"># 允许重写规则，其余一律是n</div></pre></td></tr></table></figure>
<p><img src="http://cdn.mmmxcc.cn/blog/20170504/210029356.png" alt=""></p>
<p>进入nginx配置文件再修改一下：<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">cd /usr/local/nginx/conf/vhost/</div></pre></td></tr></table></figure></p>
<p>完整配置文件应该是这个样子<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div><div class="line">8</div><div class="line">9</div><div class="line">10</div><div class="line">11</div><div class="line">12</div><div class="line">13</div><div class="line">14</div><div class="line">15</div><div class="line">16</div><div class="line">17</div><div class="line">18</div><div class="line">19</div><div class="line">20</div><div class="line">21</div><div class="line">22</div><div class="line">23</div><div class="line">24</div><div class="line">25</div><div class="line">26</div><div class="line">27</div><div class="line">28</div><div class="line">29</div><div class="line">30</div><div class="line">31</div><div class="line">32</div><div class="line">33</div></pre></td><td class="code"><pre><div class="line">server &#123;</div><div class="line"> listen 80;</div><div class="line"> #listen [::]:80;</div><div class="line">server_name ss.feiyang.li n;</div><div class="line">index index.html index.htm index.php default.html default.htm default.php;</div><div class="line">root /home/wwwroot/ss.feiyang.li/public;</div><div class="line"> </div><div class="line">location / &#123;</div><div class="line">                        try_files $uri $uri/ /index.php$is_args$args;</div><div class="line">                &#125;</div><div class="line"></div><div class="line">location ~ [^/]\.php(/|$) &#123;</div><div class="line"> # comment try_files $uri =404; to enable pathinfo</div><div class="line"> try_files $uri =404;</div><div class="line"> fastcgi_pass unix:/tmp/php-cgi.sock;</div><div class="line"> fastcgi_index index.php;</div><div class="line"> include fastcgi.conf;</div><div class="line"> #include pathinfo.conf;</div><div class="line">fastcgi_param PHP_ADMIN_VALUE &quot;open_basedir = /home/wwwroot/:/temp/:/proc&quot;;</div><div class="line">&#125;</div><div class="line"></div><div class="line"> </div><div class="line"> location ~ .*\.(gif|jpg|jpeg|png|bmp|swf)$ &#123;</div><div class="line"> expires 30d;</div><div class="line"> &#125;</div><div class="line"> </div><div class="line"> location ~ .*\.(js|css)?$</div><div class="line"> &#123;</div><div class="line"> expires 12h;</div><div class="line"> &#125;</div><div class="line"> </div><div class="line"></div><div class="line">&#125;</div></pre></td></tr></table></figure></p>
<h3 id="新建ss数据库"><a href="#新建ss数据库" class="headerlink" title="新建ss数据库"></a>新建ss数据库</h3><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div></pre></td><td class="code"><pre><div class="line">mysql -uroot -p</div><div class="line"># 输入自己的密码</div><div class="line">create database ss;</div><div class="line">use ss;</div><div class="line">source  /home/wwwroot/ss-panel/db.sql;</div></pre></td></tr></table></figure>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170504/212157512.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<p>现在访问你的站点，就可以看到 ss-panel 啦：</p>
<figure class="image-bubble">
                <div class="img-lightbox">
                    <div class="overlay"></div>
                    <img src="http://cdn.mmmxcc.cn/blog/20170505/124252257.png?imageView2/0/format/webp/interlace/1/q/75|watermark/2/text/ZmVpeWFuZy5saQ==/font/Y29taWMgc2FucyBtcw==/fontsize/500/fill/IzAzQTlGNA==/dissolve/35/gravity/SouthEast/dx/10/dy/10|imageslim
" alt="" title="">
                </div>
                <div class="image-caption"></div>
            </figure>
<h2 id="添加管理员账户"><a href="#添加管理员账户" class="headerlink" title="添加管理员账户"></a>添加管理员账户</h2><p>文字排序会让强迫症奔溃。。。。。<br><code>php xcat createAdmin</code></p>
<h2 id="邮箱设置"><a href="#邮箱设置" class="headerlink" title="邮箱设置"></a>邮箱设置</h2><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div></pre></td><td class="code"><pre><div class="line">vi .env</div><div class="line">mailgun_key = &apos;key-0****c09&apos; //上面的KEY</div><div class="line">mailgun_domain = &apos;https://api.mailgun.net/v3/feiyang.li&apos; //你经过验证过的域名</div><div class="line">mailgun_sender = &apos;postmaster@****.net&apos; //postmaster@你经过验证过的域名</div></pre></td></tr></table></figure>
<h2 id="https证书"><a href="#https证书" class="headerlink" title="https证书"></a>https证书</h2><p><code>lnmp ssl add</code><br>/home/wwwroot/ss-panel/1_ss.feiyang.li_bundle.crt<br>/home/wwwroot/ss-panel/2_ss.feiyang.li.key<br>获得绿色小锁~<br><img src="http://cdn.mmmxcc.cn/blog/20170505/124311675.png" alt=""></p>
<h1 id="安装-SS-PY-MU-服务端"><a href="#安装-SS-PY-MU-服务端" class="headerlink" title="安装 SS PY MU 服务端"></a>安装 SS PY MU 服务端</h1><h2 id="安装libsodium支持chacha20"><a href="#安装libsodium支持chacha20" class="headerlink" title="安装libsodium支持chacha20"></a>安装libsodium支持chacha20</h2><p>centos：<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div></pre></td><td class="code"><pre><div class="line">yum -y groupinstall &quot;Development Tools&quot;</div><div class="line">wget https://github.com/jedisct1/libsodium/releases/download/1.0.11/libsodium-1.0.11.tar.gz</div><div class="line">tar xf libsodium-1.0.11.tar.gz &amp;&amp; cd libsodium-1.0.11</div><div class="line">./configure &amp;&amp; make -j2 &amp;&amp; make install</div><div class="line">echo /usr/local/lib &gt; /etc/ld.so.conf.d/usr_local_lib.conf</div><div class="line">ldconfig</div></pre></td></tr></table></figure></p>
<p>ubuntu/debian：<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div></pre></td><td class="code"><pre><div class="line">apt-get install build-essential</div><div class="line">wget https://github.com/jedisct1/libsodium/releases/download/1.0.11/libsodium-1.0.11.tar.gz</div><div class="line">tar xf libsodium-1.0.11.tar.gz &amp;&amp; cd libsodium-1.0.11</div><div class="line">./configure &amp;&amp; make -j2 &amp;&amp; make install</div><div class="line">ldconfig</div></pre></td></tr></table></figure></p>
<h2 id="将代码-clone-到本地："><a href="#将代码-clone-到本地：" class="headerlink" title="将代码 clone 到本地："></a>将代码 clone 到本地：</h2><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div></pre></td><td class="code"><pre><div class="line">cd ~</div><div class="line">git clone https://github.com/fsgmhoward/shadowsocks-py-mu.git</div><div class="line">cd ~/shadowsocks-py-mu/shadowsocks/</div></pre></td></tr></table></figure>
<p>其中的 shadowsocks 子目录才是我们需要的，外面的是 setup.py 的相关文件。</p>
<h2 id="配置-shadowsocks-manyuser"><a href="#配置-shadowsocks-manyuser" class="headerlink" title="配置 shadowsocks-manyuser"></a>配置 shadowsocks-manyuser</h2><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div></pre></td><td class="code"><pre><div class="line">cp config_example.py config.py</div><div class="line">vi config.py</div></pre></td></tr></table></figure>
<p>修改其中第 15 行和第 29~31 行的内容：</p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div><div class="line">6</div><div class="line">7</div></pre></td><td class="code"><pre><div class="line"># 启用 MultiUser API</div><div class="line">API_ENABLED = True</div><div class="line"></div><div class="line"># 就是在你的站点地址后面加个 /mu</div><div class="line">API_URL = &apos;http://ss.feiyang.li/mu&apos;  </div><div class="line"># 还记得上面在 .env 中填写的 muKey 吗？把它填在这里</div><div class="line">API_PASS = &apos;api_key_just_for_test&apos;</div></pre></td></tr></table></figure>
<p>由于我们选择使用 Mu API 来与前端通信，所以我们不需要修改 config.py 中任何关于数据库的配置。</p>
<p><img src="http://cdn.mmmxcc.cn/blog/20170505/134246486.png" alt=""></p>
<p>好了，现在可以试着运行一下<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">python servers.py</div></pre></td></tr></table></figure></p>
<p>（注意，是 <strong>servers.py</strong> 而不是 server.py）。如果没错的话，应该可以看到这样的输出：</p>
<h2 id="配置-ss-manyuser-守护进程"><a href="#配置-ss-manyuser-守护进程" class="headerlink" title="配置 ss-manyuser 守护进程"></a>配置 ss-manyuser 守护进程</h2><ul>
<li><p>安装 supervisor （用的是上面安装过的 pip）：</p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div></pre></td><td class="code"><pre><div class="line">yum install python-setuptools</div><div class="line">easy_install supervisor</div></pre></td></tr></table></figure>
</li>
<li><p>创建 supervisor 配置文件</p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">echo_supervisord_conf &gt; /etc/supervisord.conf</div></pre></td></tr></table></figure>
</li>
</ul>
<ul>
<li><p>运行 supervisor 服务</p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">supervisord</div></pre></td></tr></table></figure>
</li>
<li><p>配置 supervisor 以监控 ss-manyuser 运行</p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">vim /etc/supervisord.conf</div></pre></td></tr></table></figure>
</li>
<li><p>在文件尾部添加如下内容并酌情修改：</p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div><div class="line">4</div><div class="line">5</div></pre></td><td class="code"><pre><div class="line">[program:ss-manyuser]</div><div class="line">command = python /root/shadowsocks-py-mu/shadowsocks/servers.py  </div><div class="line">user = root  </div><div class="line">autostart = true  </div><div class="line">autorestart = true</div></pre></td></tr></table></figure>
</li>
</ul>
<p>其中 command 里的目录请自行修改为你的 servers.py所在的绝对路径。</p>
<ul>
<li><p>重启 supervisor 服务以加载配置</p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">killall -HUP supervisord</div></pre></td></tr></table></figure>
</li>
<li><p>查看 shadowsocks-manyuser 是否已经运行：</p>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div></pre></td><td class="code"><pre><div class="line">ps -ef | grep servers.py</div><div class="line">root 952 739 0 15:40 ? 00:00:00 python /root/shadowsocks-rm/shadowsocks/servers.py</div></pre></td></tr></table></figure>
</li>
<li><p>可以通过以下命令管理 shadowsock-manyuser 的状态</p>
</li>
</ul>
<figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">supervisorctl &#123;start|stop|restart&#125; ss-manyuser</div></pre></td></tr></table></figure>
<h1 id="谷歌BBR加速"><a href="#谷歌BBR加速" class="headerlink" title="谷歌BBR加速"></a>谷歌BBR加速</h1><h2 id="OpenVZ-以外的（-KVM-、-Xen-、-VMware-等）"><a href="#OpenVZ-以外的（-KVM-、-Xen-、-VMware-等）" class="headerlink" title="OpenVZ 以外的（ KVM 、 Xen 、 VMware 等）"></a>OpenVZ 以外的（ KVM 、 Xen 、 VMware 等）</h2><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div><div class="line">3</div></pre></td><td class="code"><pre><div class="line">wget --no-check-certificate https://github.com/teddysun/across/raw/master/bbr.sh</div><div class="line">chmod +x bbr.sh</div><div class="line">./bbr.sh</div></pre></td></tr></table></figure>
<p>详细信息见： <a href="https://www.91yun.org/archives/5174" target="_blank" rel="external">https://www.91yun.org/archives/5174</a></p>
<h2 id="OpenVZ-小鸡"><a href="#OpenVZ-小鸡" class="headerlink" title="OpenVZ 小鸡"></a>OpenVZ 小鸡</h2><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div></pre></td><td class="code"><pre><div class="line">wget --no-check-certificate https://raw.githubusercontent.com/mmmwhy/LKL_BBR/master/lkl/install.sh &amp;&amp; bash install.sh</div></pre></td></tr></table></figure>
<p>如果运行之后，发现不能连接服务器，可能因为防火墙设置问题。可以直接关闭node节点的防火墙<br>对于CentOS:<br><figure class="highlight plain"><table><tr><td class="gutter"><pre><div class="line">1</div><div class="line">2</div></pre></td><td class="code"><pre><div class="line">service iptables stop #关闭命令：  </div><div class="line">chkconfig iptables off# 永久关闭防火墙：</div></pre></td></tr></table></figure></p>
<p>详细内容见：<a href="https://www.91yun.org/archives/6281" target="_blank" rel="external">https://www.91yun.org/archives/6281</a> </p>
<h1 id="其余常见错误："><a href="#其余常见错误：" class="headerlink" title="其余常见错误："></a>其余常见错误：</h1><p><a href="http://feiyang.li/2017/05/03/ss-panel/index.html#常见错误">http://feiyang.li/2017/05/03/ss-panel/index.html#常见错误</a></p>
<p>与上文一样，本文参考了如下博主文章，在此致谢：<br><a href="https://blessing.studio/build-shadowsocks-sharing-site-with-ss-panel/" target="_blank" rel="external">https://blessing.studio/build-shadowsocks-sharing-site-with-ss-panel/</a><br><a href="http://bitzhi.com/2016/03/install-ss-go-mu-and-ss-panel-v3/" target="_blank" rel="external">http://bitzhi.com/2016/03/install-ss-go-mu-and-ss-panel-v3/</a><br><a href="https://levyhsu.com/?p=98" target="_blank" rel="external">https://levyhsu.com/?p=98</a></p>
]]></content>
    
    <summary type="html">
    
      &lt;blockquote&gt;
&lt;p&gt;系统介绍： 1g内存，单核，也就是传说中的腾讯1元学生机。centos6.8 x32。&lt;/p&gt;
&lt;/blockquote&gt;
&lt;hr&gt;
    
    </summary>
    
      <category term="shadowsocks" scheme="http://feiyang.li/categories/shadowsocks/"/>
    
    
      <category term="GFW" scheme="http://feiyang.li/tags/GFW/"/>
    
      <category term="Linux" scheme="http://feiyang.li/tags/Linux/"/>
    
      <category term="教程" scheme="http://feiyang.li/tags/%E6%95%99%E7%A8%8B/"/>
    
      <category term="vps" scheme="http://feiyang.li/tags/vps/"/>
    
  </entry>
  
</feed>