You are here

Importance Sampling of the Transition Path Process

Event Category:
Learning Learning
Gabriel Earle

Transition path sampling is an important class of methods which seek to understand the behavior of rare but important transitions between high probability states in a dynamical system. We present a new method for dealing with this problem, based on extending results on the importance sampling of path space via the Girsanov theorem. We relate the effectiveness of the importance sampling to the quality of a solution to a related Hamilton-Jacobi-Bellman equation, and discuss how this result can inform the construction of neural networks to better facilitate the sampling of the transition path process.

Thursday, May 16, 2024 - 11:30am
LGRT 1621