Reinforcement Learning for Generating Useful Combinatorial Data

The use of AI in pure mathematics and science is a relatively young field with only a few, but significant early successes. With groundbreaking work such as IBM’s AI-Descartes and Deepmind’s Alphatensor opening the door to the possibilities of AI tools in science, I’m interested in taking seeing how far we can push these results to different areas in mathematics to see what tools might be available to mathematicians in the future. In this talk, I’ll give a brief overview of my preliminary work in applying reinforcement learning to generate useful data for mathematicians in combinatorics to inform conjectures and proofs.

You can find my slides here.