RL for Coxeter Groups
Deep Q-learning to explore the Sperner property in finite Coxeter groups
This project investigates Conjecture 1.3 on the strongly Sperner property of finite Coxeter groups using reinforcement learning. We trained a deep Q-learning agent to search for large antichains in the weak order posets of Coxeter groups of types A, B, D, and E.
The results provide computational evidence consistent with 1-Sperner behavior in D₅, D₆, and E₆ — groups where the largest antichain equals the group order.
This work was done under Prof. Oliver Pechenik at the University of Waterloo as an NSERC USRA project (Spring 2022).