Computational Visualization Center

Computational Visualization Center Wordpress Link

Semi-Supervised Variational Inference for Generative Materials Design

Today’s most significant advances in atomic-scale materials design rest on half a century of theoretical and computational developments in ab initio electronic structure techniques based on density functional theory and wave function-based methods. While these methods continue to enjoy tremendous success, predicting complex materials properties remains incredibly time-consuming even as we approach exascale computing. The main contribution of this paper is the design, training and testing of a deep learning framework based on a generative variational auto-encoder (for ab initio materials design) using FAIR principles.