Near-Optimal Primal-Dual Algorithm for Learning Linear Mixture CMDPs with Adversarial Rewards

Near-Optimal Primal-Dual Algorithm for Learning Linear Mixture CMDPs with Adversarial Rewards explores A theoretical algorithm for safe reinforcement learning in linear mixture CMDPs with adversarial rewards, achieving near-optimal regret bounds.. Commercial viability score: 3/10 in Safe Reinforcement Learning.