🤗 Datasets | 📚 Reading | 🏆 Final Results | 📝 Changelog
This repository contains the materials and code for the CSE472 (Fall 2025) course project "Beyond the Blanket Challenge".
The project is designed and mentoerd by Shu Wan.
Note
This project is finalized.
| Rank | Group | Members | Final Score (↓) |
|---|---|---|---|
| 🥇 1 | 10 | Ang, Muhammed | 0.194036 |
| 🥈 2 | 8 | Fredo, Anton | 0.208636 |
| 🥉 3 | 7 | Dhruv, Sahajpreet | 0.280585 |
| 🏅 4 | 1 | Sameera, Tanmayi | 0.304475 |
Feature selection is a critical step in machine learning, directly impacting model performance, interpretability, and generalization capabilities.
Causal Feature Selection (CFS) tackles the challenge by selecting features based on their causal relationships with the target variable. It has been shown that a set of features known as the Markov blanket, provide optimal predictability to the target. However, recent work shows that many CFS methods are outperformed by non-causal models, and struggle to generalize across domains.
Your Mission: Design a model that performs causal feature selection and achieves robust performance across different environments.
Project Duration: 7 Weeks (Oct 10–Nov 27, 2025)
| Phase | Duration | Focus Areas |
|---|---|---|
| Warm-up | Oct 10–21, 2025 | Reading / Implementation |
| Phase 1 | Oct 22–30, 2025 | Implementation / Tasks |
| Phase 2 | Oct 31–Nov 5, 2025 | Implementation / Tasks |
| Phase 3 | Nov 6–12, 2025 | Implementation / Tasks |
| Final Evaluation | Nov 13–27, 2025 | Implementation / Report |
See detailed schedule in SCHEDULE.md
I already added required environment settings in pyproject.toml. You can use uv to set up the environment:
uv venv # start a virtual environment
source .venv/bin/activate # activate the virtual environment
uv sync --all-groups # sync with pyproject.tomlconda, pip, or other environment management tools can also be used.
- 🤗 Hugging Face Datasets: our project Hugging Face space, hosting datasets, models and more
- 📚 Reading Materials: Reading materials and where to download them
- 📝 Changelog
| Deliverable | Description |
|---|---|
| Challenge Submission | Complete code, model, and reproduction instructions |
| Project Report | Comprehensive analysis (template to be provided) |
| Additional Materials | Any course-specific requirements |
If the project goes well, could students be included in a paper, or is it mainly about replication?
The project builds on an ongoing research that's still in an exploratory stage. The instructor will ensure all tasks have reasonable, verifiable answers and will be actively involved in the process. While the main goal is to explore and understand the work, there is potential for meaningful contributions depending on progress.
I'm not enrolled in the course or couldn't join a team. Can I still participate?
Yes! All materials are public—feel free to follow along and ask questions. We welcome independent learners and contributors.
Can I use LLMs for this project?
Absolutely! LLMs are encouraged for brainstorming and assistance. However, you must:
- Disclose where they were used in your final report
- Verify the correctness of any generated content
- Take full responsibility for all submitted work
Can teams discuss the project with each other?
Yes, but only discuss high-level ideas and concepts. Do not share code or specific implementation details. All submissions must be your own work
How much compute do I need?
Moderate requirements:
- Access to Sol should be sufficient
- Google offers free Colab Pro for student accounts
- Most experiments can run on standard hardware
What do winners receive?
Glory and more:
- 🏆 Glory, ☝️ Eternal bragging rights
- Potential research and publication opportunities
This project is an independent educational project created for learning and research purposes. It is not an official ASU course resource, and it is not affiliated with or endorsed by Arizona State University. All content, datasets, and code are provided as-is for public educational use.
2025 Built with ❤️.