Applied Mathematical Seminar41: Deep adaptive densitay approximation for Fokker-Planck type equations

Seminar| Institute of Mathematical Sciences

Speaker: Li Zeng, Fuzhou University

Abstract: In recent years, deep learning algorithms based on deep neural networks have been widely applied to solving high-dimensional partial differential equations, which include physics-informed neural networks (PINNs), Deep Ritz method, and so on. In this talk, we start from Fokker-Planck equations and propose flow-based adaptive sampling strategies to improve the efficiency and accuracy of PINNs for solving partial differential equations whose solutions are probability density functions.