{"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/on-policy-trust-region-policy-optimisation","title":"On-Policy Trust Region Policy Optimisation with Replay Buffers","arxiv_id":"1901.06212","date":"2019-01-18","proceeding":"ICLR 2019 5","authors":["Dmitry Kangin","Nicolas Pugeault"],"abstract":"Building upon the recent success of deep reinforcement learning methods, we\ninvestigate the possibility of on-policy reinforcement learning improvement by\nreusing the data from several consecutive policies. On-policy methods bring\nmany benefits, such as ability to evaluate each resulting policy. However, they\nusually discard all the information about the policies which existed before. In\nthis work, we propose adaptation of the replay buffer concept, borrowed from\nthe off-policy learning setting, to create the method, combining advantages of\non- and off-policy learning. To achieve this, the proposed algorithm\ngeneralises the $Q$-, value and advantage functions for data from multiple\npolicies. The method uses trust region optimisation, while avoiding some of the\ncommon problems of the algorithms such as TRPO or ACKTR: it uses\nhyperparameters to replace the trust region selection heuristics, as well as\nthe trainable covariance matrix instead of the fixed one. In many cases, the\nmethod not only improves the results comparing to the state-of-the-art trust\nregion on-policy learning algorithms such as PPO, ACKTR and TRPO, but also with\nrespect to their off-policy counterpart DDPG.","url_abs":"http://arxiv.org/abs/1901.06212v1","url_pdf":"http://arxiv.org/pdf/1901.06212v1.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":"on-policy-trust-region-policy-optimisation","repo_url":"https://github.com/dkangin/baselines","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":null},{"paper_slug":"on-policy-trust-region-policy-optimisation","repo_url":"https://github.com/dkangin/baselines/tree/master/baselines/trpo_replay","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null}],"tasks":[{"task_slug":"continuous-control","task_name":"Continuous Control"},{"task_slug":"deep-reinforcement-learning","task_name":"Deep Reinforcement Learning"},{"task_slug":"policy-gradient-methods","task_name":"Policy Gradient Methods"},{"task_slug":"reinforcement-learning","task_name":"Reinforcement Learning"},{"task_slug":"reinforcement-learning-1","task_name":"Reinforcement Learning (RL)"},{"task_slug":"reinforcement-learning-2","task_name":"reinforcement-learning"}],"methods":[{"method_slug":"adam","method_name":"Adam"},{"method_slug":"batch-normalization","method_name":"Batch Normalization"},{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"ddpg","method_name":"DDPG"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"entropy-regularization","method_name":"Entropy Regularization"},{"method_slug":"experience-replay","method_name":"Experience Replay"},{"method_slug":"ppo","method_name":"PPO"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"trpo","method_name":"TRPO"},{"method_slug":"weight-decay","method_name":"Weight Decay"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}