Neural PPO-Clip Attains Global Optimality: A Hinge Loss Perspective

Nai-Chieh Huang, Ping-Chun Hsieh, Kuo-Hao Ho +4

Policy optimization is a fundamental principle for designing reinforcement learning algorithms, and one example is the proximal policy optimization algorithm with a clipped surroga…