EXXA is a comprehensive machine learning project under ML4SCI focused on applying state-of-the-art AI techniques to exoplanet research and protoplanetary disk analysis.
This repository contains multiple Google Summer of Code (GSoC) contributions exploring various machine learning approaches for:
- Detecting exoplanets in protoplanetary disks
- Characterizing atmospheres of exoplanets
- Denoising astronomical observations
- Analyzing kinematic data from telescopes
- Applying quantum ML to exoplanet science
- Python 3.10+ (Tested with Python 3.13)
- 10+ GB free disk space
- CUDA-capable GPU (recommended, but CPU works)
# Clone the repository
git clone https://github.com/ML4SCI/EXXA.git
cd EXXA
# Run setup script (Windows)
.\setup.ps1
# Or install manually
python -m venv venv
.\venv\Scripts\Activate.ps1 # Windows
# source venv/bin/activate # Linux/Mac
pip install -r requirements.txtpython test_setup.pySee SETUP.md for detailed installation instructions.
Goal: Detect non-Keplerian features in disk observations using unsupervised learning
Tech: Transformers, Autoencoders, Domain Adaptation
Goal: Identify chemical species from exoplanet transmission spectra
Tech: CNN, LSTM, GRU, POSEIDON
Goal: Self-supervised learning on protoplanetary disk images
Tech: Masked Autoencoders, Vision Transformers
Goal: Leverage rotational symmetries in astronomical images
Tech: e2cnn, Steerable CNNs, Equivariant VGG16
EQUIVARIANT_NETWORKS_PLANETARY_SYSTEMS_ARCHITECTURES/- Best Result: 96% accuracy
Goal: Denoise ALMA/VLT observations using diffusion models
Tech: DDPM, DDIM, Diffusion Networks
DENOISING_DIFFUSION/- In Progress - Contributions Welcome!
Goal: Detect planets in dust continuum images
Tech: FARGO3D, RADMC3D, CNNs
Goal: Find exoplanets using kinematic analysis
Tech: RegNet, EfficientNetV2, PHANTOM+MCFOST
Goal: Apply QML to exoplanet characterization
Tech: Quantum Algorithms, POSEIDON
Goal: Analyze Kepler light curves
Tech: Time series analysis, Deep learning
Goal: Binary classification of TESS exoplanet candidates
Tech: CNNs, TESS data
We welcome contributions! Here's how to get started:
- Pick a project that interests you
- Read the subproject README for specific requirements
- Fork the repository and create a feature branch
- Make your changes following Python best practices
- Submit a pull request with a clear description
- Improve documentation and tutorials
- Fix bugs or add tests
- Implement new features
- Experiment with new architectures
- Add visualization tools
- Optimize performance
See individual project READMEs for specific contribution opportunities.
- Check individual project folders for relevant papers
- See
references/folder for literature
- Most projects use simulated data from FARGO3D, PHANTOM, MCFOST
- Real observations from ALMA, VLT, Kepler, TESS
- Contact mentors for dataset access
Organization: ML4SCI
Mentors: Available via ml4-sci@cern.ch
Apply: Google Form
- University of Alabama
- Oxford University
- CERN
- Questions: ml4-sci@cern.ch
- Issues: GitHub Issues
- Discussions: GitHub Discussions
See individual project folders for licensing information.
This project is part of Google Summer of Code and is supported by ML4SCI, CERN, and participating universities.
Last Updated: March 2026
Active Projects: 10
Contributors: Multiple GSoC participants