Description
Driven by a huge increase in data and compute capacity, massive machine learning systems have broken exciting new ground across a wide variety of applications. In addition to their empirical success, these systems are of significant interest from a theoretical perspective as they exhibit a number of intriguing phenomena which are hard to reconcile with conventional machine learning wisdom. Examples of such phenomena include grokking, edge of stability, double descent, benign and tempered overfitting, the generalization puzzle of highly overparametrized models and scaling laws. A number of works have emerged seeking to build a mathematical theory of these phenomena, using tools from a diverse range of disciplines including high-dimensional statistics, random matrix theory, differential equations, differential and algebraic geometry and polyhedral combinatorics to name just a few.The goal of this workshop is to facilitate further progress in developing the mathematical foundations of machine learning by considering both modern perspectives on classical questions, which have been advanced in recent years, as well as understanding emerging phenomena for which we are yet to identify appropriate mathematical frameworks. These topics are important not only for building a more complete theoretical picture, but also for spurring innovation towards new fundamental insights and improved methodologies.
This workshop will bring together a diverse group of researchers in order to shine a spotlight on recent advances, identify important open problems, share emerging ideas and tools as well as facilitate new collaborations and interdisciplinary research.
| Period | 8 Mar 2026 → 13 Mar 2026 |
|---|---|
| Event type | Workshop |
| Degree of Recognition | International |