Mathematics Colloquia and Seminars

Return to Colloquia & Seminar listing

Classifying data gathered by scattering is a ubiquitous problem, the most famous being image recognition. It has also proved to be a difficult problem, with consistent results only coming in the past 10 years due to the effectiveness of convolutional neural networks (CNNs). In this talk, I will describe one explanation for why CNNs have been so successful using the scattering transform. I will also discuss the application of the scattering transform to a specific problem: the discrimination of objects on the seabed, especially unexploded ordinance. I will also talk about a new kind of scattering transform using a shearlet transform which has specific benefits in this problem over the original variety using wavelets.

Register for pizza here.