Opening Black-Box AI Models for Constructive Mathematics

ABSTRACT

In this talk, Coco Xiaoyu Huang, Research Assistant Professor of Mathematics, Temple University, will present several collaborative projects showing how machine learning can contribute to constructive results in algebra and combinatorics.

First, although deep learning models are often treated as black boxes, we show that one can extract explicit algorithms from the computations they learn internally. In particular, we identify a new algorithmic realization of the zeta map, a classical bijection on Dyck paths. Second, Dr. Huang will discuss how reinforcement learning (RL) has been used to make progress on several conjectural problems in group theory and commutative algebra.

BIOGRAPHY

Coco Xiaoyu Huang is a Research Assistant Professor at Temple University. Dr. Huang received a Ph.D. from the City University of New York under the direction of Krzysztof Klosin. Her research interests include number theory, and the intersection between machine learning and mathematics.

A campus map is available at https://tour.stevens.edu.
Additional information is available at https://web.stevens.edu/algebraic/.