BIMSA > Seminar on Control Theory and Nonlinear Filtering Deep Learning Based Yau-Yau Algorithm for Time-Varying Filtering Problems
Deep Learning Based Yau-Yau Algorithm for Time-Varying Filtering Problems
Organizer
Shing Toung Yau
Speaker
Zeju Sun
Time
Wednesday, August 7, 2024 2:30 PM - 3:00 PM
Venue
理科楼A-304
Abstract
In this talk, I will present the recent progress on deep learning based Yau-Yau algorithm for time-varying filtering problems. A convergence analysis will first be given, with the main idea being generalized from time-invariant case. Theoretically, the deep learning based Yau-Yau algorithm can be implemented with different network structures, such as recurrent neural networks and semi-martingale driven neural stochastic differential equations. The training and implementation details of these two network structures will also be discussed.