Human-aware PGMs and Probabilistic Inference via Lifted Model Reconciliation (HAPPI)

More Information can be found here

Acknowledgments

This work was partially funded by the Ministry of Culture and Science of the German State of North Rhine-Westphalia.

Source Code Structure

• docs/ Documentation
  • build/ contains build as HTML
  • src/ Markdown files and assets
  • Project.toml Docs Project Configuration
  • make.jl build script
• instances/ input files for scripts
  • input/
  • mimic/ mimic dataset
    • .gitkeep
    • patients.csv
    • procedures_icd.csv
  • ljt-v1.0-jar-with-dependencies.jar inference algorithm
• results/ results will be put here
  • plot.r script to generate plots
  • prepare_plot.jl
• scripts/ scripts that use the hACP package
• src/ code of the hACP package
  • color_passing/
  • database/ database / dataframe / learning fg
  • fg/ factor graph
  • pfg/ paramterized fg
  • hACP.jl package file
  • helper.jl helping methods
  • hierarchical_ordering.jl
  • queries.jl
• test/ tests for the hACP Package
• Project.toml Project Configuration for hACP

When dependecies were updates:

• build docker image new: in hACP folder (login with credentials for user)

docker login registry.git.nrw
docker build --platform linux/amd64 -t registry.git.nrw/jan.speller/hacp_eacp_julia .
docker push registry.git.nrw/jan.speller/hacp_eacp_julia

Getting started

prepare hACP project

cd hACP
julia --project=.
]
instantiate

prepare docs project

cd hACP/docs
julia --project=.
]
dev ../
add Documenter

cd hACP/docs
julia --project=. make.jl

add package

cd hACP
julia --project=.
]
add package_name

Luttermann et al

Moreover, we applied openjdk version 11.0.25 to run the (lifted) inference algorithms, which are provided in the .jar file located at instances/ljt-v1.0-jar-with-dependencies.jar.

Instance Generation

First, the input instances need to be generated. To do so, see how to run at instance_generator.jl and julia mimic_generator.jl in the scripts/ directory. To be able to run mimic_generator.jl, both patients.csv and procedures_icd.csv need to be present in the instances/mimic/ directory (can be downloaded at https://physionet.org/content/mimiciv/3.1/). The input instances are then written to instances/input/ and to instances/mimic/, respectively.

Experiments

In all Scripts file there is a statement at the top line how to run those files. All results are written into the results/ directory. To create the plots, run julia prepare_plot.jl in the results/ directory to combine the obtained run times into averages and afterwards execute the R script plot.r (also in the results/ directory). The R script will then create a bunch of .tex files containing the plots of the experiments in the results/ directory. To generate the plots as .pdf files instead, set use_tikz = FALSE in line 7 of plot.r before executing the R script plot.r.