{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/mirror-descent-policy-optimization","title":"Mirror Descent Policy Optimization","arxiv_id":"2005.09814","date":"2020-05-20","proceeding":"ICLR 2022 4","authors":["Manan Tomar","Lior Shani","Yonathan Efroni","Mohammad Ghavamzadeh"],"abstract":"Mirror descent (MD), a well-known first-order method in constrained convex optimization, has recently been shown as an important tool to analyze trust-region algorithms in reinforcement learning (RL). However, there remains a considerable gap between such theoretically analyzed algorithms and the ones used in practice. Inspired by this, we propose an efficient RL algorithm, called {\\em mirror descent policy optimization} (MDPO). MDPO iteratively updates the policy by {\\em approximately} solving a trust-region problem, whose objective function consists of two terms: a linearization of the standard RL objective and a proximity term that restricts two consecutive policies to be close to each other. Each update performs this approximation by taking multiple gradient steps on this objective function. We derive {\\em on-policy} and {\\em off-policy} variants of MDPO, while emphasizing important design choices motivated by the existing theory of MD in RL. We highlight the connections between on-policy MDPO and two popular trust-region RL algorithms: TRPO and PPO, and show that explicitly enforcing the trust-region constraint is in fact {\\em not} a necessity for high performance gains in TRPO. We then show how the popular soft actor-critic (SAC) algorithm can be derived by slight modifications of off-policy MDPO. Overall, MDPO is derived from the MD principles, offers a unified approach to viewing a number of popular RL algorithms, and performs better than or on-par with TRPO, PPO, and SAC in a number of continuous control tasks. Code is available at \\url{https://github.com/manantomar/Mirror-Descent-Policy-Optimization}.","url_abs":"https://arxiv.org/abs/2005.09814v5","url_pdf":"https://arxiv.org/pdf/2005.09814v5.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"mirror-descent-policy-optimization","repo_url":"https://github.com/manantomar/Mirror-Descent-Policy-Optimization","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}}],"tasks":[{"task_slug":"continuous-control","task_name":"Continuous Control"},{"task_slug":"reinforcement-learning-1","task_name":"Reinforcement Learning (RL)"},{"task_slug":"continuous-control","task_name":"continuous-control"}],"methods":[{"method_slug":"entropy-regularization","method_name":"Entropy Regularization"},{"method_slug":"mdpo","method_name":"MDPO"},{"method_slug":"ppo","method_name":"PPO"},{"method_slug":"trpo","method_name":"TRPO"}],"datasets_introduced":[],"methods_introduced":[{"slug":"mdpo","name":"MDPO","full_name":"Mirror Descent Policy Optimization"}],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/2005.09814","atlas_url":"https://app.syntology.ai/?focus=2005.09814","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}