Analysis seminar: An IPDF talk
Wednesday, December 6 at 10.30 am
=========================
Venue: Meeting ID: 818 7750 5751, Passcode: 671479
https://us06web.zoom.us/j/81877505751?pwd=rqLYSbLv1bYxahPYpPa0cGMmNXEjPr.1
Host: Mayukh Mukherjee
Speaker: Ramesh Chandra Sau,
Affiliation: The Chinese University of Hong Kong
Title: An Analysis and Solution of Optimal Control Problems: Classical to
Modern Approaches.
Abstract: In this talk, I will present both classical approaches (e.g., EFM) and modern approaches (using deep learning tools) to solve and analyze optimal control problems. The first part of this talk will be based on the energy space formulation of Dirichlet boundary control problems. We propose a finite element-based numerical method to solve the Dirichlet boundary control problem and derive error estimates in the energy norm. In the second part, we discuss solving optimal control problems using physics-informed neural networks (C-PINN). We describe $L^2(\Omega)$ error bounds in terms of neural network parameters and number of sampling points. We present some numerical examples to illustrate the approach (C-PINN) and compare it with existing approaches