[[Tableofcontents]] = machine learning = 1. Supervised learning 1. Unsupervised learning 1. Reinforcement learning == supervised learning == * í•™ìŠµì„ ì‹œí‚¬ ë•Œ labelì— ì •ë‹µì´ ìžˆëŠ” 것 * Need input, target * Learning from difference between prediction and target * e.g. mnist, classification == unsupervised learning == * label ì´ ë¯¸ë¦¬ ì •í•´ì ¸ 있지 ì•Šì€ ê²ƒ * Need input * Cluster by distance between inputs * Can't predict outcome * e.g. clustering == reinforcement learning == * ì¼ì¢…ì˜ unsupervised learning * input : environment, reward, output : action * Learn from try * Model free * e.g. game play, stock trading == reinforcement learning == * Q learning * + Neural Network * DQN : Deep Q Learning == Basic knowledge == * MDP : Markov Decision Process * Bellman equation * Dynamic programming * Value, Polish * Value function, Polish function * Value iteration, Polish iteration == 실습 == * [https://gym.openai.com gym]: Reinforcement learningì„ ìœ„í•œ ê³ ì „ ê²Œìž„ë“¤ì„ python으로 í¬íŒ…í•œ toolkit. ì§ì ‘ 구현한 ê²ƒë„ ìžˆê³ atari는 í¬íŒ…함. * 오늘 ì‹¤ìŠµí• [https://gym.openai.com/envs/CartPole-v0 cartpole] * [https://github.com/openai/gym in github] * 필요한 ë¼ì´ë¸ŒëŸ¬ë¦¬: numpy, gym, tensorflow í•„ìš” {{{ $ pip install gym $ pip install tensorflow }}} 1. cartpole ì‹¤í–‰ì„ í•´ë³´ìž! - cartpole_init.py 1. random action(왼쪽, 오른쪽)ì„ í•˜ëŠ” cartpole - cartpole_random.py 1. q-network(q-learningì˜ NNë²„ì „) - cartpole.py 1. DQN - cartpole_dqn.py 1. 2015ì— Deep Mindì—ì„œ 발표한 DQN - cartpole_dqn2015.py == Reference == * 발표 슬ë¼ì´ë“œ: [https://slides.com/rabierre/playing_a_game_with_rl slide] * 코드: [https://github.com/Rabierre/cartpole github] * 논문: [https://arxiv.org/abs/1312.5602 Playing Atari with Deep Reinforcement Learning] == Furthermore == * [https://en.wikipedia.org/wiki/David_Silver_(programmer) David Silver]ì˜ ê°•ì˜ * [http://www0.cs.ucl.ac.uk/staff/d.silver/web/Teaching.html ê°•ì˜ë…¸íŠ¸] * [https://www.youtube.com/watch?v=2pWv7GOvuf0 ê°•ì˜ ì˜ìƒ] * Gitbook: [https://www.gitbook.com/book/dnddnjs/rl/details Fundamental of Reinforcement Learning] == í•˜ê³ ì‹¶ì€ ë§ ==