A Fresh Look at Learning Long-timescale Phenomena from Trajectory Data

Jonathan Weare, New York University

An informal pre-seminar chat with the speaker, to which all are invited, will be held at at 1:00 p.m.

Abstract: Events that occur on very long timescales are often the most interesting features of complicated dynamical systems. Even when we are able to reach the required timescales with a sufficiently accurate computer simulation, the resulting high dimensional data is difficult to mine for useful information about the event of interest. Markov state modeling (MSM) in particular has proven a powerful tool for turning trajectory data into useful understanding of long-timescale processes. Taking a new perspective on trajectory data analysis methods I will describe a family of methods that generalize MSMs with the aim of computing predictions of specific long timescale phenomena using only relatively short trajectory data (e.g. much shorter than the return time of the event). This new perspective points in exciting new directions for both rare event analysis algorithms as well mathematical analysis. In particular, I will explain the remarkable error properties in approximations of specific rare event forecasts achievable using a data set of short trajectories alone.