ABSTRACT

Many problems in applied sciences can be abstractly formulated as systems navigating over complex energy landscapes. Well-known examples include conformational changes of bio-molecules, chemical reactions, nucleation events during phase transitions, and extreme events that may lead to material or system failure, etc. Such events occur infrequently relative to the relaxation timescale of the system, yet when they do occur, they usually happen rather rapidly and have significant consequences. In this talk, I will discuss numerical methods for studying these rare events, focusing on the string method and recent developments based on machine learning techniques.