More actions
No edit summary |
(Repair MoniWiki formatting after migration) |
||
| (56 intermediate revisions by 5 users not shown) | |||
| Line 1: | Line 1: | ||
__TOC__ | |||
= machine learning = | = machine learning = | ||
머신 러닝의 세가지 분류 | |||
# Supervised learning | |||
# Unsupervised learning | |||
# Reinforcement learning | |||
== supervised learning == | == supervised learning == | ||
* 학습을 시킬 때 | * 학습을 시킬 때 input으로 feature(입력값)와 label(원하는 결과값)을 함께 전달 | ||
* Learning from difference between prediction and label | |||
* Learning from difference between prediction and | |||
* e.g. mnist, classification | * e.g. mnist, classification | ||
== unsupervised learning == | == unsupervised learning == | ||
* | * input: feature만 입력, 보통 projection등으로 feature의 차원을 축소시킨다. | ||
* Cluster by distance between inputs | * Cluster by distance between inputs | ||
* | * Human can't predict the outcome | ||
* e.g. clustering | * e.g. clustering | ||
== reinforcement learning == | == reinforcement learning == | ||
| Line 17: | Line 18: | ||
* input : environment, reward, output : action | * input : environment, reward, output : action | ||
* Learn from try | * Learn from try | ||
* Model free | ** Model free: 게임의 규칙을 알려주지 않음 | ||
* e.g. game play, stock trading | * e.g. game play, stock trading | ||
== reinforcement learning == | == reinforcement learning == | ||
* Q learning | * Q learning | ||
* Neural Network | * Q learning + Neural Network | ||
* DQN : Deep Q Learning | * DQN : Deep Q Learning | ||
** hidden layer를 늘리는게 다가 아니다! | |||
== Basic knowledge == | == Basic knowledge == | ||
* MDP : Markov Decision Process | * MDP : Markov Decision Process | ||
| Line 32: | Line 34: | ||
== 실습 == | == 실습 == | ||
numpy, gym, tensorflow 필요 | * [https://gym.openai.com gym]: Reinforcement learning을 위한 고전 게임들을 python으로 포팅한 toolkit. 직접 구현한 것도 있고 atari는 포팅함. [https://github.com/openai/gym github]에 코드가 공개되어 있다. | ||
** 오늘 실습할 [https://gym.openai.com/envs/CartPole-v0 cartpole] | |||
* 필요한 라이브러리: numpy, gym, tensorflow 필요 | |||
$ pip install gym | |||
$ pip install tensorflow | |||
=== 순서 === | |||
# 일단 cartpole 실행을 해보자! - [https://github.com/Rabierre/cartpole/blob/master/cartpole_init.py cartpole_init.py] | |||
# random action(왼쪽, 오른쪽)을 하는 cartpole - [https://github.com/Rabierre/cartpole/blob/master/cartpole_random.py cartpole_random.py] | |||
== | # q-network(q-learning의 NN버전) - [https://github.com/Rabierre/cartpole/blob/master/cartpole_qnetwork.py cartpole_qnetwork.py] | ||
# DQN - [https://github.com/Rabierre/cartpole/blob/master/cartpole_dqn.py cartpole_dqn.py] | |||
# 2015에 Deep Mind에서 발표한 DQN - [https://github.com/Rabierre/cartpole/blob/master/cartpole_dqn2015.py cartpole_dqn2015.py] | |||
== | == Reference == | ||
* 발표 슬라이드: [https://slides.com/rabierre/playing_a_game_with_rl slide] | |||
* 실습코드: [https://github.com/Rabierre/cartpole github] | |||
* DeepMind의 DQN 논문: [https://arxiv.org/abs/1312.5602 Playing Atari with Deep Reinforcement Learning] | |||
* Tensorflow tutorial: [https://github.com/golbin/TensorFlow-Tutorials/tree/master/10%20-%20DQN DQN] | |||
== Furthermore == | |||
* [https://en.wikipedia.org/wiki/David_Silver_(programmer) David Silver]의 강의 | |||
** [http://www0.cs.ucl.ac.uk/staff/d.silver/web/Teaching.html 강의노트] | |||
** [https://www.youtube.com/watch?v=2pWv7GOvuf0 강의 영상] | |||
* Gitbook: [https://www.gitbook.com/book/dnddnjs/rl/details Fundamental of Reinforcement Learning]. 한글로 되어 있다! | |||
* 레퍼런스 모음: [[Machine Learning]] | |||
== 후기 및 기타의견 == | |||
Latest revision as of 00:34, 29 March 2026
machine learning
머신 러닝의 세가지 분류
- Supervised learning
- Unsupervised learning
- Reinforcement learning
supervised learning
- 학습을 시킬 때 input으로 feature(입력값)와 label(원하는 결과값)을 함께 전달
- Learning from difference between prediction and label
- e.g. mnist, classification
unsupervised learning
- input: feature만 입력, 보통 projection등으로 feature의 차원을 축소시킨다.
- Cluster by distance between inputs
- Human can't predict the outcome
- e.g. clustering
reinforcement learning
- 일종의 unsupervised learning
- input : environment, reward, output : action
- Learn from try
- Model free: 게임의 규칙을 알려주지 않음
- e.g. game play, stock trading
reinforcement learning
- Q learning
- Q learning + Neural Network
- DQN : Deep Q Learning
- hidden layer를 늘리는게 다가 아니다!
Basic knowledge
- MDP : Markov Decision Process
- Bellman equation
- Dynamic programming
- Value, Polish
- Value function, Polish function
- Value iteration, Polish iteration
실습
- gym: Reinforcement learning을 위한 고전 게임들을 python으로 포팅한 toolkit. 직접 구현한 것도 있고 atari는 포팅함. github에 코드가 공개되어 있다.
- 오늘 실습할 cartpole
- 필요한 라이브러리: numpy, gym, tensorflow 필요
$ pip install gym $ pip install tensorflow
순서
- 일단 cartpole 실행을 해보자! - cartpole_init.py
- random action(왼쪽, 오른쪽)을 하는 cartpole - cartpole_random.py
- q-network(q-learning의 NN버전) - cartpole_qnetwork.py
- DQN - cartpole_dqn.py
- 2015에 Deep Mind에서 발표한 DQN - cartpole_dqn2015.py
Reference
- 발표 슬라이드: slide
- 실습코드: github
- DeepMind의 DQN 논문: Playing Atari with Deep Reinforcement Learning
- Tensorflow tutorial: DQN
Furthermore
- David Silver의 강의
- Gitbook: Fundamental of Reinforcement Learning. 한글로 되어 있다!
- 레퍼런스 모음: Machine Learning