| 書(shū)目名稱(chēng) | Reinforcement Learning |
| 編輯 | Richard S. Sutton |
| 視頻video | http://file.papertrans.cn/826/825930/825930.mp4 |
| 叢書(shū)名稱(chēng) | The Springer International Series in Engineering and Computer Science |
| 圖書(shū)封面 |  |
| 描述 | Reinforcement learning is the learning of a mapping fromsituations to actions so as to maximize a scalar reward orreinforcement signal. The learner is not told which action to take, asin most forms of machine learning, but instead must discover whichactions yield the highest reward by trying them. In the mostinteresting and challenging cases, actions may affect not only theimmediate reward, but also the next situation, and through that allsubsequent rewards. These two characteristics -- trial-and-errorsearch and delayed reward -- are the most important distinguishingfeatures of reinforcement learning. .Reinforcement learning is both a new and a very old topic in AI. Theterm appears to have been coined by Minsk (1961), and independently incontrol theory by Walz and Fu (1965). The earliest machine learningresearch now viewed as directly relevant was Samuel‘s (1959) checkerplayer, which used temporal-difference learning to manage delayedreward much as it is used today. Of course learning and reinforcementhave been studied in psychology for almost a century, and that workhas had a very strong impact on the AI/engineering work. One could infact consider all of reinforcement learning to |
| 出版日期 | Book 1992 |
| 關(guān)鍵詞 | agents; algorithms; artificial intelligence; control; learning; machine learning; proving; reinforcement le |
| 版次 | 1 |
| doi | https://doi.org/10.1007/978-1-4615-3618-5 |
| isbn_softcover | 978-1-4613-6608-9 |
| isbn_ebook | 978-1-4615-3618-5Series ISSN 0893-3405 |
| issn_series | 0893-3405 |
| copyright | Springer Science+Business Media New York 1992 |
|Archiver|手機(jī)版|小黑屋|
派博傳思國(guó)際
( 京公網(wǎng)安備110108008328)
GMT+8, 2026-1-23 19:14
發(fā)表于 2025-3-21 16:35:22
