Artificial Intelligence and Data-driven Materials Exploration

Yongqing Cai (蔡永青)*

Institute of Applied Physics and Materials Engineering, University of Macau

ABSTRACT: Artificial intelligence represented by data-driven method has attracted worldwide attention with the increase in data volume and the development of computer computing power. It has brought about a revolution in many research fields, repeatedly breaking state of the art records in image processing, speech recognition and natural language processing. Recent years have witnessed great progress of application of artificial intelligence in materials science. My report attempts to offer an overview of the recent successful data-driven materials research strategies, introduce these researches and demonstrate challenges that materials informatics is facing. The materials science community is just beginning to explore and utilize the plethora of available informatic algorithms to mine and learn from big data. From the prospective of machine learning, data-driven strategies are composed of two distinct steps. One step is to represent numerically the various input cases in dataset, each input case would been processed to a string of numbers after this step. Another is establishing a mapping between input and target properties, several algorithms are available to establish this mapping such as linear regression, decision tree, SVM, neural networks, even ensemble learning. In my report, I will give a description in these two steps to demonstrate the process of data-driven method establishment for materials informatics. Furthermore, I will introduce several typical works on materials science using machine learning method.

Keywords：Artificial intelligence; Materials Informatics; Machine Learning