When
1 – 2 p.m., Feb. 19, 2025
Where
Speaker: Andrew Arnold, Program in Applied Mathematics
Title: GBO for PDEs: Computing gradients in PDE-constrained optimization
Abstract: Knowing the local gradient of an objective function with respect to design variables is imperative for the construction of sample-efficient optimizers. How does one obtain the gradient in the case that the objective function contains terms that are defined through PDEs? In this tutorial, I make the analogy between training artificial neural networks through automatic differentiation and gradient-based PDE-constrained optimization, with an application to topological optimization of an elastic material.