If I'm training an agent using SARSA and the epsilon greedy mechanism to encourage exploration, how can I test/evaluate the learned policy?

Can I simply run the algorithm with epsilon set to zero so that the algorithm always selects the greedy action and evaluate the learned policy that way, or is this method no longer SARSA?

I am keeping epsilon constant, so as far as I understand, my SARSA is learning a near-optimal policy.

New contributor
james_erikson is a new contributor to this site. Take care in asking for clarification, commenting, and answering. Check out our Code of Conduct.

Your Answer

james_erikson is a new contributor. Be nice, and check out our Code of Conduct.

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.