Please join us for a UW Data Science Seminar featuring Donsub Rim from the Mathematics Department at Washington University in St. Louis on Wednesday, May 28th from 4:30 to 5:20 p.m. PT. The seminar will be held in Electrical and Computer Engineering Building 125 – Campus Map.

“A Stability of Neural Networks and Its Applications to Tsunami Early Warning”

Abstract: Feedforward neural networks (NNs) have remarkable potential to serve as fast and accurate prediction models in many high-consequence applications. However, current NN models are not yet suitable for deployment in practice, since they are known to be unstable with respect to input perturbations called adversarial examples. The focus of this talk is in analyzing these instabilities. We will introduce low rank Householder expansion (LRHE) of NNs, a certain linearization of the NN about an input, and discuss computational experiments illustrating a close relationship between the low-rank structure revealed by LRHE and the adversarial examples. Throughout this talk, we will focus on a particular geophysical application in tsunami early warning, namely NN models that aim to predict tsunami waveforms solely from geodetic measurements of the earthquake.

This a joint work with Sanah Suri (Wash U), Sanghyun Hong (Oregon State U), Kookjin Lee (Arizona State U), and Randall J. LeVeque (U Washington).

Biography: Donsub Rim is an Assistant Professor of Mathematics in the Mathematics Department at Washington University in St. Louis. He received his PhD at the University of Washington, and held postdoctoral positions at Columbia University and New York University. His current research interests are in dimensionality reduction of solutions of nonlinear wave equations, analysis of neural networks and tsunami early warning.

The 2024-2025 seminars will be held in person, and are free and open to the public.