{"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/exploration-a-study-of-count-based","title":"#Exploration: A Study of Count-Based Exploration for Deep Reinforcement Learning","arxiv_id":"1611.04717","date":"2016-11-15","proceeding":"NeurIPS 2017 12","authors":["Haoran Tang","Rein Houthooft","Davis Foote","Adam Stooke","Xi Chen","Yan Duan","John Schulman","Filip De Turck","Pieter Abbeel"],"abstract":"Count-based exploration algorithms are known to perform near-optimally when\nused in conjunction with tabular reinforcement learning (RL) methods for\nsolving small discrete Markov decision processes (MDPs). It is generally\nthought that count-based methods cannot be applied in high-dimensional state\nspaces, since most states will only occur once. Recent deep RL exploration\nstrategies are able to deal with high-dimensional continuous state spaces\nthrough complex heuristics, often relying on optimism in the face of\nuncertainty or intrinsic motivation. In this work, we describe a surprising\nfinding: a simple generalization of the classic count-based approach can reach\nnear state-of-the-art performance on various high-dimensional and/or continuous\ndeep RL benchmarks. States are mapped to hash codes, which allows to count\ntheir occurrences with a hash table. These counts are then used to compute a\nreward bonus according to the classic count-based exploration theory. We find\nthat simple hash functions can achieve surprisingly good results on many\nchallenging tasks. Furthermore, we show that a domain-dependent learned hash\ncode may further improve these results. Detailed analysis reveals important\naspects of a good hash function: 1) having appropriate granularity and 2)\nencoding information relevant to solving the MDP. This exploration strategy\nachieves near state-of-the-art performance on both continuous control tasks and\nAtari 2600 games, hence providing a simple yet powerful baseline for solving\nMDPs that require considerable exploration.","url_abs":"http://arxiv.org/abs/1611.04717v3","url_pdf":"http://arxiv.org/pdf/1611.04717v3.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":"exploration-a-study-of-count-based","repo_url":"https://github.com/clementbernardd/Count-Based-Exploration","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"exploration-a-study-of-count-based","repo_url":"https://github.com/nhynes/abc","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}},{"paper_slug":"exploration-a-study-of-count-based","repo_url":"https://github.com/uoe-agents/derl","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"atari-games","task_name":"Atari Games"},{"task_slug":"continuous-control","task_name":"Continuous Control"},{"task_slug":"deep-reinforcement-learning","task_name":"Deep Reinforcement Learning"},{"task_slug":"reinforcement-learning","task_name":"Reinforcement Learning"},{"task_slug":"reinforcement-learning-1","task_name":"Reinforcement Learning (RL)"},{"task_slug":"continuous-control","task_name":"continuous-control"},{"task_slug":"reinforcement-learning-2","task_name":"reinforcement-learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/atari-games-on-atari-2600-freeway","task":"Atari Games","dataset":"Atari 2600 Freeway","model":"TRPO-hash","rank_in_archive_order":1,"of":59,"metrics":{"Score":"34.0"},"uses_additional_data":false},{"leaderboard":"/sota/atari-games-on-atari-2600-frostbite","task":"Atari Games","dataset":"Atari 2600 Frostbite","model":"TRPO-hash","rank_in_archive_order":15,"of":53,"metrics":{"Score":"5214.0"},"uses_additional_data":false},{"leaderboard":"/sota/atari-games-on-atari-2600-montezumas-revenge","task":"Atari Games","dataset":"Atari 2600 Montezuma's Revenge","model":"TRPO-hash","rank_in_archive_order":29,"of":50,"metrics":{"Score":"75"},"uses_additional_data":false},{"leaderboard":"/sota/atari-games-on-atari-2600-venture","task":"Atari Games","dataset":"Atari 2600 Venture","model":"TRPO-hash","rank_in_archive_order":23,"of":55,"metrics":{"Score":"445.0"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1611.04717","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}