机器学习在高分子材料基因组研究中的进展与挑战

Machine learning (ML) plays an important role in the investigation and development of polymer material genomes. The success of ML-based studies strongly depends on the design and selection of feature descriptors, which reasonably portray chemical and structural characteristics of polymer materials. In this review, we elucidate a few of descriptors commonly utilized for effectively constructing the link between polymer structures, chemical compositions and aggregate structures, macroscopic properties. In addition, the database, especial for the polymer materials, is also explicitly listed, although the continuous development of specific database is still in a large demand. The research progress of ML methods in the field of polymer materials in recent years is reviewed, as well as successful applications and achievements. In particular, the solutions to deal with the small amount of data or the high cost of expensive data are also presented. According to the current research progress, the difficulty and challenge of ML applications in the field of polymer materials are discussed as well.