{"id":"arxiv:2305.19452","title":"Bigger, Better, Faster: Human-level Atari with human-level efficiency","date":"2023-05-30","section":"Reinforcement learning","quality":0.228173,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/","source_url":"https://arxiv.org/abs/2305.19452","text_available":true,"authors":["Max Schwarzer","Johan Obando-Ceron","Aaron Courville","Marc Bellemare"],"outgoing":[{"source":"arxiv:2305.19452","target":"arxiv:1509.06461","evidence":"id","id":"arxiv:1509.06461","title":"Deep Reinforcement Learning with Double Q-learning","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/a2f115ac879cd2efb9548baf3ecdcb925d2492f03b4b0940720f6e5b0840715b/"},{"source":"arxiv:2305.19452","target":"arxiv:1511.05952","evidence":"title","id":"arxiv:1511.05952","title":"Prioritized Experience Replay","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/c88bf0c11c1817dcc457c001dfef5cd120a0946c8a009a0fe40648dae03eb7d4/"},{"source":"arxiv:2305.19452","target":"arxiv:1511.06581","evidence":"title","id":"arxiv:1511.06581","title":"Dueling Network Architectures for Deep Reinforcement Learning","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/0a5f0593cf943c782ba38ae2fc386f953f859b9d3fc9722e8f17f67ca9cf50b5/"},{"source":"arxiv:2305.19452","target":"arxiv:1707.06887","evidence":"title","id":"arxiv:1707.06887","title":"A Distributional Perspective on Reinforcement Learning","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/96b040bb771d28ca3254b5519a2191c3e958a35dfdd60075344ac86af6a590e1/"},{"source":"arxiv:2305.19452","target":"arxiv:1710.02298","evidence":"id","id":"arxiv:1710.02298","title":"Rainbow: Combining Improvements in Deep Reinforcement Learning","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4183a5aecdb6558733bd0b2028e6197904c8b46173181bf6c8270fb6cbcfc973/"},{"source":"arxiv:2305.19452","target":"arxiv:2005.01643","evidence":"id","id":"arxiv:2005.01643","title":"Offline Reinforcement Learning: Tutorial, Review, and Perspectives on Open Problems","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/2ae228e6f0a418b4cd46f033b57c7452ba84a9c8cb985b198224186333c17d97/"},{"source":"arxiv:2305.19452","target":"arxiv:2011.14826","evidence":"title","id":"arxiv:2011.14826","title":"Revisiting Rainbow: Promoting more Insightful and Inclusive Deep Reinforcement Learning Research","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/2bf20803a5d9e686543fe5f3967e69b5aa481ba29ff91c9ca3fbeb147ab8c0aa/"},{"source":"arxiv:2305.19452","target":"arxiv:2205.07802","evidence":"title","id":"arxiv:2205.07802","title":"The Primacy Bias in Deep Reinforcement Learning","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4e2aca370b941bb16bb6fefc359c420f6d4bde5a956bedaf0a84e6e71febc5bc/"},{"source":"arxiv:2305.19452","target":"arxiv:2301.04104","evidence":"id","id":"arxiv:2301.04104","title":"Mastering Diverse Domains through World Models","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/266d59e63d2f705efdc2eb7ad93f5126b7092b4450efd0d8c17099c51574b176/"},{"source":"arxiv:2305.19452","target":"openreview:OpC-9aBBVJe","evidence":"url","id":"openreview:OpC-9aBBVJe","title":"Sample-Efficient RL by Breaking the Replay Ratio Barrier (ICLR 2023, precursor of BBF)","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/8bab1f9ee93a0613f60e3e9577b9ddcad6937e59c156709d0351478c38f1d5bf/"},{"source":"arxiv:2305.19452","target":"url:minedojo.org","evidence":"title","id":"url:minedojo.org","title":"MineDojo: Building Open-Ended Embodied Agents with Internet-Scale Knowledge","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/0a1b263777b8aec5e510481b45b7b4e11d45d7c127fd41dc37b5ff908f6420d2/"}],"incoming":[{"source":"arxiv:2411.03820","target":"arxiv:2305.19452","evidence":"title","id":"arxiv:2411.03820","title":"Beyond The Rainbow: High Performance Deep Reinforcement Learning on a Desktop PC","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/8bcc96ef608c017d3ddb7ded7dc9e3b760390f7df441214eeee08a2586c981ea/"},{"source":"arxiv:2503.14858","target":"arxiv:2305.19452","evidence":"title","id":"arxiv:2503.14858","title":"1000 Layer Networks for Self-Supervised RL: Scaling Depth Can Enable New Goal-Reaching Capabilities","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/dfa04045239215a4a49544cd31c178090abc3a587e98bd9af08e0ea96d658911/"},{"source":"doi:10.1038/s41586-024-07711-7","target":"arxiv:2305.19452","evidence":"title","id":"doi:10.1038/s41586-024-07711-7","title":"Maintaining Plasticity in Deep Continual Learning","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/b5d46327415fe38f2b9160b874735530939961a71536f7f1ac4c0dc7a3af3224/"}],"text_parts":[{"start":0,"end":15973,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-1.html"},{"start":15973,"end":31924,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-2.html"},{"start":31924,"end":32110,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-3.html"}],"contents":[{"title":"Bigger, Better, Faster: Human-level Atari with human-level efficiency","offset":0,"level":1,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-1.html#text"},{"title":"Abstract","offset":73,"level":6,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-1.html#text"},{"title":"Keywords:","offset":619,"level":6,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-1.html#text"},{"title":"1 Introduction","offset":637,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-1.html#text"},{"title":"2 Background","offset":4027,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-1.html#text"},{"title":"3 Related Work","offset":9485,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-1.html#text"},{"title":"4 Method","offset":13163,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-1.html#text"},{"title":"Base agent.","offset":14500,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-1.html#text"},{"title":"Harder resets.","offset":15302,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-1.html#text"},{"title":"Larger network.","offset":15952,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-1.html#text"},{"title":"Receding update horizon.","offset":17245,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-2.html#text"},{"title":"Increasing discount factor.","offset":18380,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-2.html#text"},{"title":"Weight decay.","offset":19108,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-2.html#text"},{"title":"Removing noisy nets.","offset":19572,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-2.html#text"},{"title":"5 Analysis","offset":20698,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-2.html#text"},{"title":"The importance of self-supervision.","offset":20833,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-2.html#text"},{"title":"Sample efficiency via more gradient steps.","offset":21522,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-2.html#text"},{"title":"The surprising importance of target networks","offset":22678,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-2.html#text"},{"title":"Reset Strength","offset":23306,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-2.html#text"},{"title":"Scale is not enough on its own.","offset":23829,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-2.html#text"},{"title":"Computational efficiency.","offset":24276,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-2.html#text"},{"title":"6 Revisiting the Atari 100k benchmark","offset":24893,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-2.html#text"},{"title":"New Frontiers","offset":26869,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-2.html#text"},{"title":"Data Scaling","offset":27333,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-2.html#text"},{"title":"7 Discussion and Future Work","offset":28766,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-2.html#text"},{"title":"Appendix A Additional Results","offset":31112,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-2.html#text"},{"title":"Instructions for reporting errors","offset":31146,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/part-2.html#text"}],"text_code_points":32110,"source_markdown":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/727a7ee43fd005401d34d4591a2f49d85175f9366ebf2c9e7a862b340787614e/source.md"}