| Abstract: This week, Sultan Hassan will be presenting "Similarity of neural network representation as a way to generalize models to out-of-distribution samples".
MLCT is a weekly seminar hosted by the Physics department where anyone—students, postdocs, faculty—can present a 45-minute-ish lecture or discussion on machine learning or statistical inference. The setting is informal, chalkboard-focused (with optional slides), and interruptions are encouraged. Topics can range from rough ideas to finished research, with the assumption that the audience is at the level of a graduate student who has some knowledge of how typical ML methods work.
To see a weekly schedule of seminars, see https://dwh.gg/MLCT |