Dry [summary] | Deep Reinforcement Learning depth of reinforcement learning

        In machine learning, we often classified as supervised learning and unsupervised learning, but try ignores an important branch of reinforcement learning. Supervised learning and unsupervised learning to distinguish very well, learning objectives, with or without labels are case standard. If the goal is to predict the supervised learning, reinforcement learning is decision making so that the surrounding environment by constantly updating status, given rewards or punishment measures to constantly adjust and give the new strategy. In short, like a child you should not eat snack time eating snacks, your mom know you will make the punishment, then the next time it will not make the same mistake, if you follow the rules, then perhaps your mother will give you some incentive, the ultimate goal is you want to eat at the dinner, when to eat snacks to eat snacks, rather than snacking at inappropriate times. Similarly, Flappy bird has swept over a period of time, a lot of players in a short time to reach the high score, is how to do it? In addition to very powerful players are really playing their own hand high, in fact, a lot of high scores through reinforcement learning method we use to train a model, learn how to let the birds do not fly straight ahead and encountered obstacles, the highest Minute. In addition, the well-known Alpha Go, actually strengthening the model of learning and training, but the depth of reinforcement learning.

 

　　After 2013 DeepMind published a Playing Atari with Deep Reinforcement Learning article, the depth of reinforcement learning it slowly into people's vision. Later, in 2015, DeepMind has published a Human Level Control through Deep Reinforcement Learning, so that the depth of intensive study to get a lot of attention, then the emergence of many academic achievements. We are familiar with the depth of reinforcement learning should be in the 16 to 17 years time, especially after the Alpha Go occur, many companies and a large number of researchers have begun to focus on the depth of reinforcement learning, and try to apply it in various application scenarios.

 

　　About the depth of reinforcement learning, I compiled some information, if interested can learn about:

 

【paper】

• "Depth of reinforcement learning" in English (150), is the last version (70) of the enhanced version: https://arxiv.org/abs/1810.06339
• Classic books: Reinforcement Learning: An Introduction (2nd Edition)
• Proceedings, broad coverage, requires a certain basis: Reinforcement Learning: State-of-Art-at The
• Two very full of paper data collection:
• GitHub - junhyukoh/deep-reinforcement-learning-papers: A list of recent papers regarding deep reinforcement learning
• GitHub - muupan/deep-reinforcement-learning-papers: A list of papers and resources dedicated to deep reinforcement learning

 

[Blog]

• yuxili:  https://medium.com/@yuxili
• Guest Post (Part I): Demystifying Deep Reinforcement Learning
• Guest Post (Part II): Deep Reinforcement Learning with Neon
• Blog Post (Part III): Deep Reinforcement Learning with OpenAI Gym
• Andrej Karpathy blog: Deep Reinforcement Learning: Pong from Pixels

 

【article】

• Nanjing University, Dr. Yang Yu: Introduction reinforcement learning (learning to strengthen the full description) https://www.leiphone.com/news/201705/uO8nd09EnR77NBRP.html

 

【course】

• Zero-based entry: Mo trouble Python: https://morvanzhou.github.io/tutorials/machine-learning/reinforcement-learning/
• David Silver enhanced learning programs (video and ppt), 2015 years, it requires a certain basis:  http://www0.cs.ucl.ac.uk/staff/D.Silver/web/Teaching.html
• The best enhanced learning materials, can be combined with David Silver of course watching: Sutton & Barto Book: Reinforcement Learning: An Introduction
• Stanford CS234: HTTP: // web.stanford.edu/class/cs234/index.html
• Berkeley CS294: http://rll.berkeley.edu/deeprlcourse/
• Pieter Abbeel of AI programs (including reinforcement learning, using Pacman experiment): Artificial Intelligence
• Pieter Abbeel depth of reinforcement learning courses: CS 294 Deep Reinforcement Learning, Fall 2015
• Nando de Freitas的深度学习课程 （有视频有ppt有作业）：Machine Learning
• Michael Littman的增强学习课程：https://www.udacity.com/course/reinforcement-learning–ud600
• 最新机器人专题课程Penn（2016年开课）：Specialization 
• Deep Learning Summer School：pptsvideos

 

 

【学习网站】

• openAI GYM Reinforcement Learning toolkits: https://gym.openai.com
• 强化学习示例演示：https://qqiang00.github.io/reinforce/javascript/demo_iteration.html
• karpathy的各种强化学习的演示：https://cs.stanford.edu/people/karpathy/reinforcejs/index.html
• MIT的强化学习在线学习网站：http://web.mst.edu/~gosavia/rl_website.html

 

【Github】

• Awesome-RL: https://github.com/aikorea/awesome-rl
• Flappybird：https://github.com/yenchenlin/DeepLearningFlappyBird
• Deep Reinforcement Learning in Tensorflow:https://github.com/carpedm20/deep-rl-tensorflow
• https://github.com/ShangtongZhang/reinforcement-learning-an-introduction
• GitHub - songrotek/DeepTerrainRL: terrain-adaptive locomotion skills using deep reinforcement learning
• GitHub - songrotek/async-rl: An attempt to reproduce the results of "Asynchronous Methods for Deep Reinforcement Learning" (http://arxiv.org/abs/1602.01783)
• GitHub - songrotek/rllab: rllab is a framework for developing and evaluating reinforcement learning algorithms.
• GitHub - songrotek/DRL-FlappyBird: Playing Flappy Bird Using Deep Reinforcement Learning (Based on Deep Q Learning DQN using Tensorflow)
• GitHub - songrotek/DeepMind-Atari-Deep-Q-Learner: The original code from the DeepMind article + my tweaks

 

 

【会议】

• NIPS 2015 Deep Reinforcement Learning Workshop
• ICLR 2016
• RSS 2016 Deep Learning for Robotics