Artificial Intelligence (AI) Mastering Development

Why Q-Learning and SARSA have terrible performance?

I am trying to solve a MDP problem with almost 50 states and 60 actions with Q-Learning or SARSA. However, the performance of both algorithms is terrible and cannot find the optimal policy found by backward induction. Does it mean that these algorithms do not converge to the optimal policies or there are some other problems? I run the algorithms for 100k, 500k and 1m episodes but I could not find the optimal policy.

Leave a Reply

Your email address will not be published. Required fields are marked *