{"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/lenient-multi-agent-deep-reinforcement","title":"Lenient Multi-Agent Deep Reinforcement Learning","arxiv_id":"1707.04402","date":"2017-07-14","proceeding":null,"authors":["Gregory Palmer","Karl Tuyls","Daan Bloembergen","Rahul Savani"],"abstract":"Much of the success of single agent deep reinforcement learning (DRL) in\nrecent years can be attributed to the use of experience replay memories (ERM),\nwhich allow Deep Q-Networks (DQNs) to be trained efficiently through sampling\nstored state transitions. However, care is required when using ERMs for\nmulti-agent deep reinforcement learning (MA-DRL), as stored transitions can\nbecome outdated because agents update their policies in parallel [11]. In this\nwork we apply leniency [23] to MA-DRL. Lenient agents map state-action pairs to\ndecaying temperature values that control the amount of leniency applied towards\nnegative policy updates that are sampled from the ERM. This introduces optimism\nin the value-function update, and has been shown to facilitate cooperation in\ntabular fully-cooperative multi-agent reinforcement learning problems. We\nevaluate our Lenient-DQN (LDQN) empirically against the related Hysteretic-DQN\n(HDQN) algorithm [22] as well as a modified version we call scheduled-HDQN,\nthat uses average reward learning near terminal states. Evaluations take place\nin extended variations of the Coordinated Multi-Agent Object Transportation\nProblem (CMOTP) [8] which include fully-cooperative sub-tasks and stochastic\nrewards. We find that LDQN agents are more likely to converge to the optimal\npolicy in a stochastic reward CMOTP compared to standard and scheduled-HDQN\nagents.","url_abs":"http://arxiv.org/abs/1707.04402v2","url_pdf":"http://arxiv.org/pdf/1707.04402v2.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":"lenient-multi-agent-deep-reinforcement","repo_url":"https://github.com/gjp1203/nui_in_madrl","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"GPL-3.0"}}],"tasks":[{"task_slug":"deep-reinforcement-learning","task_name":"Deep Reinforcement Learning"},{"task_slug":"multi-agent-reinforcement-learning","task_name":"Multi-agent Reinforcement Learning"},{"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":"experience-replay","method_name":"Experience Replay"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1707.04402","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}