2020/11/04 by Rohit Batra, Batra, Rohit, Hanjun Dai +13 · 3 citations
Computer Science · Materials Science · #FOS: Computer and information sciences #FOS: Physical sciences #Machine Learning (cs.LG) #Machine Learning and Algorithms #Machine Learning in Materials Science #Soft Condensed Matter (cond-mat.soft)
paper · pdf · doi:10.48550/arxiv.2011.02551
openalex publication_date 2020/11/04 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28
The design/discovery of new materials is highly non-trivial owing to the\nnear-infinite possibilities of material candidates, and multiple required\nproperty/performance objectives. Thus, machine learning tools are now commonly\nemployed to virtually screen material candidates with desired properties by\nlearning a theoretical mapping from material-to-property space, referred to as\nthe \forward problem. However, this approach is inefficient, and severely\nconstrained by the candidates that human imagination can conceive. Thus, in\nthis work on polymers, we tackle the materials discovery challenge by solving\nthe \inverse problem: directly generating candidates that satisfy desired\nproperty/performance objectives. We utilize syntax-directed variational\nautoencoders (VAE) in tandem with Gaussian process regression (GPR) models to\ndiscover polymers expected to be robust under three extreme conditions: (1)\nhigh temperatures, (2) high electric field, and (3) high temperature \and\nhigh electric field, useful for critical structural, electrical and energy\nstorage applications. This approach to learn from (and augment) human ingenuity\nis general, and can be extended to discover polymers with other targeted\nproperties and performance measures.\n