AI RESEARCH
Last-Iterate Convergence of General Parameterized Policies in Constrained MDPs
arXiv CS.AI
•
ArXi:2408.11513v2 Announce Type: replace-cross This paper focuses on learning a Constrained Marko Decision Process (CMDP) via general parameterized policies. We propose a Primal-Dual based Regularized Accelerated Natural Policy Gradient (PDR-ANPG) algorithm that uses entropy and quadratic regularizers to reach this goal.