Machine learning meets physics: bridging the gap between big and small data

Lei-Han TANG (Professor, Physics)

Current ICTS team members: Lei-Han TANG (Leader), Pan-Jun KIM, Liang TIAN, Xian YANG, Zhong-Ying ZHAO

Big data and machine-learning algorithms have fundamentally changed social sciences and sectors of the health industry. To achieve the same type of transformative impact in physical sciences, i.e., biology, chemistry and physics, while subject to the long-established tradition and rigor of these disciplines, a deeper understanding of what current generation of AI algorithms can and cannot do is required. Progress in this area will have enormous influence on how we formulate research projects big or small, all the way to improving the efficiency of day-to-day execution at the bench or in front of the computer.

Model development is a powerful methodology that allows integration of fundamental physical principles with emerging data and observations. In the language of machine learning, this means that we need to combine the general learning algorithms with rule-based modelling to extract maximum understanding and predictive power under limited data. How much data is sufficient for hands-off computation or in other words, when do we need to seek input from experts or dive into the literature to reformulate the objective function and the ML architecture? These questions have been actively explored in the statistical physics community, leveraging on the idea of energy landscapes, response and fluctuations, and the renormalisation group.

Seminar on "Materials Informatics: The 4th paradigm" on 11 Sept 2018


ICTS Workshop on "Architecture of Deep Neural Networks" on 14 Jan 2018