Researchers at IIT Kharagpur, along with the Indo-Korea Science and Technology Center (IKST), have developed a deep-learning framework, CrysXPP, that will allow for rapid and accurate prediction of electronic, magnetic, and elastic properties of a wide range of materials.
In a recently published paper in NPJ Computational Materials (a journal of the Nature Publishing Group), the researchers said that CrysXPP “lowers the need for large property tagged datasets by intelligently designing an autoencoder, CrysAE.”
CrysXPP: An Explainable Property Predictor for CrystallineMaterials – YouTube
The paper added that the crucial structural and chemical properties captured by CrysAE from a large amount of available crystal graphs data helped in achieving low prediction errors. The researchers have also designed a feature selector that helps to interpret the model’s prediction. “When given a small amount of experimental data, CrysXPP is consistently able to outperform conventional DFT,” added the paper.
Image: CrysXPP: An explainable property predictor for crystalline materials
Pawan Goyal, associate professor in computer science and engineering, IIT Kharagpur, and fellow researcher of this study, in an interaction with the KGP Chronicle, said that in terms of the future, the team is planning to undertake a larger-scale study using more materials. They are also planning to use the predictor as a reward function to accelerate the generation of new materials.
Read the full paper here.
