Title: Identifying equivalent Calabi-Yau topologies: a discrete challenge from mathematics and physics for machine learning

Speaker: Washington Taylor (MIT)

Time: 14:45-15:20 (GMT+0), 2021.08.10

Abstract:

Calabi-Yau manifolds are a class of geometric spaces that are of central interest in physics and mathematics. Hundreds of millions of distinct constructions of these manifolds are known, but there is no systematic way of confirming when two constructions give the same manifold. In particular, there is no known finite time algorithm for checking the equivalence of the simple topological data of two such constructions encoded in the "triple intersection" form, which is essentially a 3-index analogue of a matrix of integers.

This talk will give an overview of this problem and some efforts towards developing machine learning algorithms to check equivalence, which leads to fundamental questions about solving discrete math problems using machine learning.