Robustness and risk management via distributional dynamic programming

Mastane Achab, Gergely Neu

In dynamic programming (DP) and reinforcement learning (RL), an agent learns to act optimally in terms of expected long-term return by sequentially interacting with its environment…