Title: The Graph Scattering Transform(s)
Abstract: The Scattering Transform is a wavelet-based model of Convolutional Neural Networks (CNNs) introduced by S. Mallat to help us better understand the great empirical success of CNNs. In recent years, several independent constructions have introduced new versions of the scattering transform for graph-structured data. In my talk, I will overview these constructions, their common themes and their difference. I will also introduce a new construction which unifies several of the previous approaches.
Time: June 17, 2020, 2:00pm – 2:30pm (eastern)