Predicting Bandgap of Hexanary Oxides (CGCNN)
2025A materials ML project where I trained and evaluated a CGCNN-based model to predict bandgaps of hexanary oxides. The focus was building a clean workflow from structure data to model training and evaluation.
What I did
- Prepared structure datasets and organized train/validation/test splits.
- Implemented a reproducible training/evaluation pipeline for CGCNN experiments.
- Tracked results to compare model performance across settings.
- Especially for Low-Bandgap region, I have enhanced its prediction through importing weight function on its code skeleton.
Why it matters
Bandgap is a key screening property for photocatalyst candidates. This project helped me practice ML workflows in materials: data handling, training loops, and evaluation discipline.
Figure
How to run
git clone https://github.com/Gibeom-KIM-02/Predicting-bandgap-of-Hexanary-oxide
cd Predicting-bandgap-of-Hexanary-oxideSee the repository README for environment setup and training commands.