Algebraic Geometry: A Window to Machine Learning

Hosted by the Institute for Pure and Applied Mathematics (IPAM) at UCLA, this workshop runs from 1 to 5 February 2027 in Los Angeles. It brings together mathematicians and machine-learning researchers to develop new mathematical frameworks for understanding learning systems, treating machine learning as the core subject rather than an applied afterthought.

The programme targets fundamental phenomena in modern deep learning where underlying algebraic or geometric structure remains poorly understood, including feature learning, delayed generalisation (grokking), neural collapse and low-rank adaptation. A poster session is part of the schedule, giving early-career participants a chance to present.

The organising committee comprises Yulia Alexandr (UCLA), Guido Montúfar (UCLA), Michael Murray (University of Bath) and Rishi Sonthalia (Boston College). Attendance is via an application and registration process linked from the official page; as with most IPAM workshops, prospective participants should check the page for funding and travel-support details. The event will appeal to researchers working at the interface of algebraic geometry, statistics and deep-learning theory.