Date/Time Date(s) - 26/04/20243:30 pm - 4:30 pm
Location: HH 305
Speaker: Samuel Cohen (University of Oxford)
Title: Approximating PDEs with wide neural networks
Abstract: Neural networks are a rich family of function approximations, which perform particularly well in high dimension. In this talk, we will consider training neural networks as approximations of the solutions of PDEs, using the ‘deep Galerkin’ and ‘Q-PDE’ algorithms. We will see that, under fairly general conditions, in the limiting regime where the network becomes infinitely wide, this approximation trained with gradient flow methods converges to a flow in an appropriate Sobolev space, and hence converges to the true solution of the PDE.
Based on joint work with Justin Sirignano and Deqing Jiang.
Coffee and cookies will be served in HH 216 at 3pm – All are welcome.
See this talk posted, here