CrysXPP: An Explainable Property Predictor for Crystalline Materials

Accepted in NPJ Computational Materials (Nature) Journal,2022. (Impact Factor: 13.9)
Kishalay Das, Bidisha Samanta, Pawan Goyal, Seung-Cheol Lee, Satadeep Bhattacharjee and Niloy Ganguly

Abstract -

We present a deep-learning framework, CrysXPP, to allow rapid and accurate prediction of electronic, magnetic, and elastic properties of a wide range of materials. CrysXPP lowers the need for large property tagged datasets by intelligently designing an autoencoder, CrysAE. The important structural and chemical properties captured by CrysAE from the large amount of available crystal graphs data helped in achieving low prediction errors. Moreover, we design a feature selector that provides interpretability to the results obtained. Most notably, when given a small amount of experimental data, CrysXPP is consistently able to outperform conventional DFT. A detailed ablation study establishes the importance of different design steps. We release the large pre-trained model CrysAE. We believe by fine-tuning the model with a small amount of property-tagged data, researchers can achieve superior performance on various applications with a restricted data source.

Model Overview -

Our proposed model has two modules :

The following paper describes the details of the CrysXPP framework: CrysXPP: An Explainable Property Predictor for Crystalline Materials

[Code] [Slides] [Video]


Cite as -

If youare using CrysXPP, please cite our work as follow :

  title={CrysXPP: An explainable property predictor for crystalline materials},
  author={Das, Kishalay and Samanta, Bidisha and Goyal, Pawan and Lee, Seung-Cheol and Bhattacharjee, Satadeep and Ganguly, Niloy},
  journal={npj Computational Materials},
  publisher={Nature Publishing Group}