Sample-Efficient Constrained Reinforcement Learning with General Parameterized Policies

Join a technical lecture where Dr. Washim Uddin Mondal, Assistant Professor at IIT Kanpur, explores sample-efficient algorithms for Constrained Markov Decision Processes (CMDP) with general parameterized policies. Discover how the proposed Primal-Dual Accelerated Natural Policy Gradient (PD-ANPG) algorithm achieves epsilon optimality and constraint violation with improved sample complexity. Learn about applications in infinite state spaces, neural network policies, and how this approach advances the field by closing gaps between theoretical bounds. Dr. Mondal brings extensive expertise from his postdoctoral research at Purdue University, academic achievements including the Prime Minister's Research Fellowship, and practical experience consulting for the Indian Army on national interest projects.