Fall 2025

Sep 5 – Robert Nowak (University of Wisconsin-Madison)
Function Space Perspectives on Neural Networks

Sep 12 – Shay Moran (Technion)
Characterizations of Uniform Learnability: Vapnik-Chervonenkis, Natarajan, and Daniely-Shalev Shwartz Dimensions

Sep 19 – Benjamin Recht (University of California, Berkeley)
The Irrational Decision

Sep 26 – No Seminar

Oct 3 – Yuejie Chi (Yale University)
Transformers Learn Generalizable Chain-of-Thought Reasoning via Gradient Descent

Oct 10 – Weijie Su (University of Pennsylvania)
Do Large Language Models (Really) Need Statistical Foundations?

Oct 17– Navid Azizan (MIT)
Hard-Constrained Neural Networks

Oct 24 – Stefan Wager (Stanford University)
Learning to Price Electricity for Optimal Demand Response

Oct 31 – Vardan Papyan (University of Toronto)
Attention Sinks: A ‘Catch, Tag, Release’ Mechanism for Embeddings

Nov 7 – Tatsunori Hashimoto (Stanford University)
Back to the future – data efficient language modeling

Nov 14 – Sewoong Oh (University of Washington)
Private statistical estimation via robustness and stability

Nov 21 – Christos Thrampoulidis (University of British Columbia)
The Implicit Geometry of Deep Representations: Insights From Log-Bilinear Softmax Models