Talk "Optimality-preserving reductions of (Neural) Differential Equations"

by Max Tschaikowski

Mon, 24 April 2023 at 04:30 pm in Paris, France

Joint work with: Luca Cardelli, Kim G. Larsen, Mirco Tribastone, Andrea Vandin and others

Abstract: In this talk I will provide an overview of my research. We will start from the efficient algorithmic reduction of large-scale systems of nonlinear ordinary differential equations (ODEs), a ubiquitous specification and modelling formalism in systems biology, engineering and performance analysis. Thereafter, we will consider optimality-preserving reduction of parameterized ODEs, formally known as optimal control problems. The talk will finish with ongoing work concerning the reduction of neural ODEs, a class of neural networks that arise as the mathematical limit of infinitely deep neural networks.