Machine Learning in Materials Science: Status and Prospects

Abstract

Dramatic improvements in hardware and algorithmic techniques have advanced computing performance, which has opened 'data-driven' opportunities in materials science. The Materials Genome Initiative (MGI), launched by the US government in 2011, has accelerated materials design by changing the underlying philosophy. Under political and technological circumstances, researchers can now successfully employ the 'rational data-driven design' strategy in their work rather than using the former Edisonian-style approach. This article provides an overview of a recent success story, the Polymer Genome Project, in which the emerging machine learning method was actively utilized. The discussion also points towards some challenge, that might be encountered in the future and the advanced strategies to overcome them.