{"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/momentum-based-accelerated-q-learning","title":"Momentum-based Accelerated Q-learning","arxiv_id":"1910.11673","date":"2019-10-23","proceeding":null,"authors":[],"abstract":"This paper studies accelerated algorithms for Q-learning. We propose an\nacceleration scheme by incorporating the historical iterates of the Q-function.\nThe idea is conceptually inspired by the momentum-based acceleration methods in\nthe optimization theory. Under finite state-action space settings, the proposed\naccelerated Q-learning algorithm provably converges to the global optimum with\na rate of $\\mathcal{O}(1/\\sqrt{T})$. While sharing a comparable theoretic\nconvergence rate with the existing Speedy Q-learning (SpeedyQ) algorithm, we\nnumerically show that the proposed algorithm outperforms SpeedyQ via playing\nthe FrozenLake grid world game. Furthermore, we generalize the acceleration\nscheme to the continuous state-action space case where function approximation\nof the Q-function is necessary. In this case, the algorithms are validated\nusing commonly adopted testing problems in reinforcement learning, including\ntwo discrete-time linear quadratic regulation (LQR) problems from the Deepmind\nControl Suite, and the Atari 2600 games. Simulation results show that the\nproposed accelerated algorithms can improve the convergence performance\ncompared with the vanilla Q-learning algorithm.","url_abs":"http://arxiv.org/abs/1910.11673v1","url_pdf":"http://arxiv.org/pdf/1910.11673v1.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":"momentum-based-accelerated-q-learning","repo_url":"https://github.com/fapont/hackaton-hiparis-2021","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"atari-games","task_name":"Atari Games"},{"task_slug":"q-learning","task_name":"Q-Learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}