Information Theory, Machine Learning and Statistics Seminar

Spring 2021

This term the Information Theory, Machine Learning and Statistics Seminar focuses on generaliza-
tion in machine learning and divergence measures, some of the topics studied last term. The seminar
consists of a series of two-hours talks aimed to graduate students and researchers with a basic know-
ledge of information theory and machine learning (e.g., as covered in the mini-courses last term).

The seminar runs in a bi-weekly format, taking place every other Friday from 9am to 11am (CDT).
The seminar sessions are transmitted using Zoom, please request the access link to Mario Diaz.

List of Speakers

Date Speaker Title
05/March Mario Diaz
Universidad Nacional Autónoma de México
Generalization in Machine Learning via (Conditional) Mutual Information
19/March Tyler Sypherd
Arizona State University
Synthesizing Classification-Calibration and Rademacher Complexity Generalization with alpha-loss
16/April Mahdi Haghifam
University of Toronto
Sharpened Generalization Bounds based on Conditional Mutual Information and an Application to Noisy, Iterative Algorithms
30/April Borja Rodríguez
KTH Royal Institute of Technology
Tighter Expected Generalization Error Bounds via Wasserstein Distance
14/May James Melbourne
On Discrete Analogs of the Entropy Power Inequality
28/May Alvaro Díaz
Max Planck Institute for the Physics of Complex Systems
A Phase Transition in Terms of the Shannon Entropy

Last update: May 22, 2021