데블스캠프2017/강화학습으로컴퓨터에게고전게임플레이시키기 (rev. 1.33)
1. machine learning ¶
- Supervised learning
- Unsupervised learning
- Reinforcement learning
1.1. supervised learning ¶
- 학습을 시킬 때 label에 정답이 있는 것
- Need input, target
- Learning from difference between prediction and target
- e.g. mnist, classification
1.2. unsupervised learning ¶
- label 이 미리 정해져 있지 않은 것
- Need input
- Cluster by distance between inputs
- Can't predict outcome
- e.g. clustering
1.3. reinforcement learning ¶
- 일종의 unsupervised learning
- input : environment, reward, output : action
- Learn from try
- Model free
- e.g. game play, stock trading
1.4. reinforcement learning ¶
- Q learning
- + Neural Network
- DQN : Deep Q Learning
1.5. Basic knowledge ¶
- MDP : Markov Decision Process
- Bellman equation
- Dynamic programming
- Value, Polish
- Value function, Polish function
- Value iteration, Polish iteration
- 필요한 라이브러리: numpy, gym, tensorflow 필요
- gym: Reinforcement learning을 위한 고전 게임들을 python으로 포팅한 toolkit. 직접 구현한 것도 있고 atari는 포팅함.
$ pip install gym
$ pip install tensorflow
- cartpole 실행을 해보자! - cartpole_init.py
- random action(왼쪽, 오른쪽)을 하는 cartpole - cartpole_random.py
- q-network(q-learning의 NN버전) - cartpole.py
- DQN - cartpole_dqn.py
- 2015에 Deep Mind에서 발표한 DQN - cartpole_dqn2015.py
