CrysAtom: Distributed Representation of Atoms for Crystal Property Prediction

This work is accepted at the Third Learning on Graphs Conference (LoG 2024)

Environment Setup

The dependency pakages can be installed using the command.

pip install -r requirements.txt

Dataset

A small subset of dataset present in the data directory. To get the datasets used for creating dense vector is present in [Zenodo]

Dense Vector Extraction

Our 200-dimensional vector is located in the dense_vector directory. To use this dense vector representation in your model, please adjust the input dimension of our GNN model accordingly. If you use our vector, kindly cite our paper. To get the dense vector representation from scratch please run the AtomVectorExtractor.py

Downstream Property Prediction

After getting the 200 dimensional feature vector from the CrysAtom, Please visit Downstream Property Predictor Task folder for more detail

Citation

Please cite our paper if it's helpful to you in your research.

@inproceedings{
mukherjee2024crysatom,
title={CrysAtom: Distributed Representation of Atoms for Crystal Property Prediction},
author={Shrimon Mukherjee and Madhusudan Ghosh and Partha Basuchowdhuri},
booktitle={The Third Learning on Graphs Conference},
year={2024},
url={https://openreview.net/forum?id=2AIVM5pWXz}
}