按年被提到
20132014 · 10 次2026
强化学习与智能体›博弈与游戏 AI·arXiv:1312.5602 · cs.LG

Playing Atari with Deep Reinforcement Learning

用深度强化学习玩 Atari 游戏
Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Alex Graves, Ioannis Antonoglou, Daan Wierstra, Martin Riedmiller · 2013

卷积网络加 Q 学习,从原始像素学会玩 Atari

开始阅读 ↗PDFalphaXiv 讨论
被推荐 · 20 人首次 2014-01-13 · 最近 2024-01-21
20不同的人
15活跃月份
18讨论串
9,458被引(OpenAlex)

提出首个能直接从高维感知输入成功学习控制策略的深度学习模型:用 Q-learning 变体训练卷积神经网络,输入是原始像素,输出是估计未来回报的价值函数。在 Arcade Learning Environment 的七款 Atari 游戏上,同一架构和算法不作调整,在六款上超过此前所有方法,其中三款超过人类专家。

HN 为什么提它

评论者把它当作 DeepMind 深度强化学习的起点:同一个系统只看像素和得分信号就能学会多款游戏,是 AlphaGo 之前 AI 玩游戏的代表性进展。

奠基之作难度 3 / 5约 30 分钟经得起时间
读之前最好知道
1强化学习与 Q-learning
2卷积神经网络
3马尔可夫决策过程(Markov decision process)
4经验回放(experience replay)

HN 书评

推荐在前,同类按时间 · 每篇最多保留 20 条,统计用全部提及
强烈推荐modeless · 2015-01

“If you want to see the state of the art in using AI to play video games, look no further than "Playing Atari with Deep Reinforcement Learning" [2], where a single general AI system learns to play many different games.”

强烈推荐,称它是 AI 玩电子游戏的最先进水平,单一系统学会多种游戏
Mario Lives: An AI for creating a living and conversing agent [video]原评论 ↗
推荐cyorir · 2014-07

“the readme links a cool paper on applying convolutional neural networks to old games.”

推荐这篇用卷积神经网络玩老游戏的酷论文
Neural networks in JavaScript原评论 ↗
推荐sinwave · 2014-12

“While this complaint generally has validity, their paper [1] does IMO present an advance; it's not just handing a bunch of labeled data off to a large neural network.”

认为它确有技术进步,不只是把标注数据丢给大神经网络
Google's Intelligence Designer原评论 ↗
引用karpathy · 2014-01

“There is a paper on Arxiv that describes the approach”

指出 arXiv 上有论文描述了这种方法
Google to Buy Artificial Intelligence Startup DeepMind for $400M原评论 ↗
引用amaks · 2014-01

“Also, they have published ArXiv paper on the subject: Playing Atari with Deep Reinforcement Learning.”

补充说他们在 arXiv 上发表了相关论文
Google to Buy Artificial Intelligence Startup DeepMind for $400M原评论 ↗
引用jkldotio · 2014-02

“which is supposedly just an algorithm working on the raw pixels and then beating all other programs and half of the human experts.[1]”

引用它说明算法只看原始像素就击败其他程序和半数人类专家
Why Watson and Siri Are Not Real AI原评论 ↗
看全部 16 条 →
引用Sniffnoy · 2014-10

“A note -- if you're linking to arXiv, it's better to link to the abstract rather than directly to the PDF.”

提醒链接 arXiv 时最好指向摘要页而不是直接指向 PDF
Playing Atari with Deep Reinforcement Learning [pdf]原评论 ↗
引用amoruso · 2014-12

“Here's the original research paper if you're interested.”

给出原始研究论文,供感兴趣的人阅读
Google's Intelligence Designer原评论 ↗
引用javierluraschi · 2015-02

“DeepMind has a paper describing this approach: Playing Atari with Deep Reinforcement Learning”

指出 DeepMind 有论文描述了这种做法
Show HN: AI QLearning Robot Simulation in JavaScript and WebGL原评论 ↗
引用daveguy · 2016-03

“The score was separate, or rather a signal representing increased score.”

说明得分是单独的信号,用来补充说明智能体怎么学
DeepMind Challenges for StarCraft原评论 ↗
引用jkldotio · 2016-05

“they use four frames in sequence. That was at the end of 2013.”

指出它用连续四帧作为输入,那是 2013 年底的事
End to End Learning for Self-Driving Cars [pdf]原评论 ↗
引用ktta · 2016-06

“Deepmind has been focussing on this since the beginning[0] and actually has great progress as far as Atari games go[1].”

拿它说明 DeepMind 在 Atari 游戏上进展很大
OpenAI technical goals原评论 ↗
引用visarga · 2016-08

“Before AlphaGo, DeepMind released a reinforcement learning algorithm that could play many Atari games just from the raw pixels on the screen, in many games surpassing humans.”

回顾 AlphaGo 之前 DeepMind 的 RL 算法,只看原始像素就在多款 Atari 游戏上超过人类
The Difference Between AI, Machine Learning, and Deep Learning原评论 ↗
引用komaromy · 2017-01

“They don't mention this in the paper, but the network wasn't trained to convergence at each training step”

补充论文没提到的细节:网络并非每个训练步都训练到收敛
CS 294 Deep Reinforcement Learning, Spring 2017原评论 ↗
引用ap46 · 2017-04

“Third paragraph of Intro in original DeepMind paper Playing Atari with Deep Reinforcement Learning”

指出某项目的介绍抄自这篇 DeepMind 论文引言第三段
Deep Reinforcement Learning to Play Space Invaders [pdf]原评论 ↗
引用gary_0 · 2024-01

“some of DeepMind's earliest work was a deep reinforcement learning model that could play seven different Atari games with no adjustments needed to switch games[2]”

引用 DeepMind 早期成果:同一个深度 RL 模型玩七款 Atari 游戏无需调整
Self-Rewarding Language Models原评论 ↗
被提交为 story · 4 次