Mathematical Sciences Department Numerical Methods Seminar
Yao Li, UMass Amherst
Thursday, April 10th
11:00am - 12:00 pm
Title: Data-driven computational methods for stochastic dynamics
Abstract: In this talk, I will summarize our recent progress in using data-driven methods to numerically study the properties of stochastic differential equations. This includes (1) Solving both time-dependent Fokker-Planck equations and stationary Fokker-Planck equations, (2) Estimating the speed of convergence to the steady states, (3) Computing forward and backward eigenfunction of the Fokker-Planck equations, and (4) Solving the Freidlin-Wentzell quasi-potential function. Compared with traditional methods, our data-driven approaches are significantly more applicable to higher dimensional problems.