• Home

  • Documents

  • Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology...
101
Reinforcement Learning Opensource to Baselines

Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

  • Upload
    others

  • View
    3

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Reinforcement LearningOpensource to Baselines

Page 2: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Member

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

2

● 한국생산기술연구원○ 로보틱스 및 재활로봇 연구

● KAIST 연구원○ IoT 시스템 개발 & 강화학습

연구

● 플랜아이 연구원○ 추천 시스템 & 강화학습 연구

● 한양대학교 석사과정○ 강화학습 시뮬레이션 구축○ 속도 예측 연구

● 호서대학교 로봇 자동화공학과 학부생○ 임베디드, ROS, 제어, 강화학습○ 강화학습을 로봇에 적용하는

연구

Page 3: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Index

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

3

● Reinforcement Learning

● Policy based Reinforcement Learning

● Value based Reinforcement Learning

● Famous open source for Reinforcement Learning

● Why?

● Requirements

● Supported algorithms

● To do List

● How to use

● License

Page 4: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Reinforcement Learning

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

4

Page 5: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Policy base Reinforcement Learning

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

5

Page 6: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Policy base Reinforcement Learning

Page 7: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Probability to select action(a) at state(s)

Policy base Reinforcement Learning

Page 8: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Policy base Reinforcement Learning

Page 9: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

State Distribution

Policy base Reinforcement Learning

Page 10: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

State Distribution

Policy

Policy base Reinforcement Learning

Page 11: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

State Distribution

Policy

Reward

Policy base Reinforcement Learning

Page 12: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Policy base Reinforcement Learning

Page 13: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Policy base Reinforcement Learning

Page 14: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Policy base Reinforcement Learning

Page 15: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Policy base Reinforcement Learning

Page 16: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Policy base Reinforcement Learning

Page 17: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Policy base Reinforcement Learning

Page 18: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Policy base Reinforcement Learning

Page 19: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Policy base Reinforcement Learning

Page 20: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Policy base Reinforcement Learning

● Policy Gradient

● Advantage Actor Critic

● Natural Policy Gradient

● Trust Region Policy Gradient

● Proximal Policy Gradient

Page 21: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Policy Gradient

● Policy Gradient Methods for Reinforcement Learning with Function Approximation

○ Policy can be trained by iterable method

○ Policy can be trained by Policy Gradient

Method

● Policy Gradient

○ Optimal Policy can be obtained by Gradient Ascend

Method

● Iterable Method

○ Optimal Policy Can be obtained by iterable method

like deep learning method

Page 22: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Policy Gradient

Page 23: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

● Why advantage is needed?

○ Reduce variance, without changing expectation

○ a

○ laksdfasdf

○

Advantage Actor Critic

Page 24: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Natural Policy Gradient

● The parameter update can not guarantee

the performance improvement of the actual function

Page 25: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Trust Region Policy Optimization● Trust Region

Page 26: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Trust Region Policy Optimization● Trust Region

Page 27: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Trust Region Policy Optimization● Trust Region

Page 28: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Trust Region Policy Optimization● Trust Region

Page 29: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Trust Region Policy Optimization● Trust Region

Page 30: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Trust Region Policy Optimization● Trust Region

Page 31: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Trust Region Policy Optimization● Trust Region

Page 32: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Trust Region Policy Optimization● Trust Region

Page 33: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Trust Region Policy Optimization● Trust Region

Page 34: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Trust Region Policy Optimization● Trust Region

Page 35: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Trust Region Policy Optimization● Trust Region

Page 36: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Proximal Policy Optimization● Trust Region

Page 37: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Proximal Policy Optimization● Trust Region

Page 38: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Proximal Policy Optimization● Trust Region

Page 39: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Proximal Policy Optimization● Trust Region

Page 40: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Proximal Policy Optimization● Trust Region

Page 41: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Proximal Policy Optimization● Trust Region

Page 42: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Proximal Policy Optimization● Trust Region

Page 43: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Proximal Policy Optimization● Trust Region

Page 44: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Proximal Policy Optimization● Trust Region

Page 45: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Proximal Policy Optimization● Trust Region

Page 46: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Proximal Policy Optimization● Trust Region

Page 47: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Proximal Policy Optimization● Trust Region

Page 48: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Proximal Policy Optimization● Trust Region

Page 49: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Proximal Policy Optimization● Trust Region

Page 50: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Value base Reinforcement Learning

Page 51: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Value base Reinforcement Learning

Page 52: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

● Playing Atari with Deep Reinforcement Learning(NIPS 2013)

● Human-level control through deep reinforcement learning(Nature 2015)

● Deep Reinforcement Learning with Double Q-Learning

● Dueling Network Architectures for Deep Reinforcement Learning

● Prioritized Experience Replay

● …..

Value base Reinforcement Learning

Page 53: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Dispatch Time : 5 min

Distance : 2 min

Value base Reinforcement Learning

Page 54: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Dispatch Time : 5 min

Distance : 2 min

Value base Reinforcement Learning

Page 55: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

3 min 8 min

Value base Reinforcement Learning

Page 56: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

● A Distributional perspective on Reinforcement Learning(2017)

● Distributional Reinforcement Learning with Quantile Regression(2017)

● Implicit Quantile Networks for Distributional Reinforcement Learning(2018)

Value base Reinforcement Learning

Page 57: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Value base Reinforcement Learning

Page 58: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Value base Reinforcement Learning

Page 59: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Distributional Reinforcement Learningwith Quantile Regrssion

● Histogram cannot meet condition

● Wasserstein Distance

Page 60: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Distributional Reinforcement Learningwith Quantile Regrssion

Page 61: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Implicit Quantile Networkfor Distributional Reinforcement Learning

Page 62: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Implicit Quantile Networkfor Distributional Reinforcement Learning

RandomSampling

Page 63: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Value base Reinforcement Learning

Page 64: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Famous open source for Reinforcement Learning

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

64

Page 65: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Famous open source for Reinforcement Learning

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

65

● Deepminds

Page 66: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Famous open source for Reinforcement Learning

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

66

● Deepminds

● Open AI

Page 67: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Famous open source for Reinforcement Learning

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

67

● Deepminds

● Open AI

Page 68: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Famous open source for Reinforcement Learning

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

68

● Deepminds

● Open AI

Page 69: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Famous open source for Reinforcement Learning

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

69

● Deepminds

● Open AI

Page 70: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Famous open source for Reinforcement Learning

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

70

● Deepminds

● Open AI

Page 71: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Famous open source for Reinforcement Learning

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

71

● Deepminds

● Open AI

Page 72: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Famous open source for Reinforcement Learning

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

72

● Deepminds

○ Dopamine

■ Value based Reinforcement Learning opensource baseline

■ https://github.com/google/dopamine

● Open AI

Page 73: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Famous open source for Reinforcement Learning

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

73

● Deepminds

○ Dopamine

■ Value based Reinforcement Learning open source baseline

■ https://github.com/google/dopamine

● Open AI

○ Baselines

■ Policy based Reinforcement Learning open source baseline

■ https://github.com/openai/baselines

Page 74: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Why?

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

74

Page 75: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Why?

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

75

● Hard to Install

Page 76: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Why?

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

76

● Hard to Install

○ baselines : mp4py, mujoco…

○ dopamine : easy

Page 77: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Why?

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

77

● Hard to Install

○ baselines : mp4py, mujoco…

○ dopamine : easy

Page 78: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Why?

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

78

● Hard to Install

○ baselines : mp4py, mujoco…

○ dopamine : easy

● Hard to use to your environment

Page 79: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Why?

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

79

● Hard to Install

○ baselines : mp4py, mujoco…

○ dopamine : easy

● Hard to use to your environment

○ manipulate the api

○ https://github.com/google/dopamine

○ https://github.com/openai/baselines

https://github.com/google/dopamine
https://github.com/openai/baselines
Page 80: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Why?

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

80

● Hard to Install

○ baselines : mp4py, mujoco…

○ dopamine : easy

● Hard to use to your environment

○ manipulate the api

○ https://github.com/google/dopamine

○ https://github.com/openai/baselines

● Korean

https://github.com/google/dopamine
https://github.com/openai/baselines
Page 81: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Requirements

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

81

Page 82: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Requirements

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

82

● Hard to Install

Page 83: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Requirements

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

83

● Easy to Install

Page 84: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Requirements

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

84

● Easy to Install

Page 85: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Requirements

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

85

● Easy to Install

● Hard to use to your environment

Page 86: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Requirements

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

86

● Easy to Install

● Easy to use to your environment

Page 87: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Requirements

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

87

● Easy to Install

● Easy to use to your environment

○ Provide many tutorial code(discrete, continuous action)

Page 88: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Requirements

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

88

● Easy to Install

● Easy to use to your environment

○ Provide many tutorial code(discrete, continuous action)

○ Keep the flow of other reinforcement learning code

Page 89: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Requirements

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

89

● Easy to Install

● Easy to use to your environment

○ Provide many tutorial code(discrete, continuous action)

○ Keep the flow of other reinforcement learning code

Page 90: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Requirements

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

90

● Easy to Install

● Easy to use to your environment

● Korean

Page 91: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Supported algorithms

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

91

Page 92: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Supported algorithms

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

92

● Vanilla Policy Gradient

● Advantage Actor Critic

● Proximal Policy Optimization

● Deep Deterministic Policy Gradient

Page 93: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Supported algorithms

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

93

● Vanilla Policy Gradient(Discrete Action)

● Advantage Actor Critic(Discrete Action)

● Proximal Policy Optimization(Discrete, Continuous Action)

● Deep Deterministic Policy Gradient(Discrete, Continuous Action)

Page 94: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

To do List

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

94

● Vanilla Policy Gradient(Discrete Action)

● Advantage Actor Critic(Discrete Action)

● Proximal Policy Optimization(Discrete, Continuous Action)

● Deep Deterministic Policy Gradient(Discrete, Continuous Action)

Page 95: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

To do List

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

95

● Vanilla Policy Gradient(Discrete Action)

● Advantage Actor Critic(Discrete Action)

● Proximal Policy Optimization(Discrete, Continuous Action)

● Deep Deterministic Policy Gradient(Discrete, Continuous Action)

● Applicate LSTM to Vanilla Policy Gradient, Advantage Actor Critic, Proximal Policy Optimization

● Actor Critic Experience Replay

● Soft actor critic

● Value based Reinforcement Learning

○ DQN, Double DQN, Deuling DQN to Rainbow

○ Distributional RL

Page 96: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

How to use

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

96

Page 97: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

How to use

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

97

Page 98: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

License

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

98

Page 99: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

License

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

99

Free

Page 100: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

100

We are hiring

Page 101: Reinforcement Learning Opensource to Baselines...Korea Advanced Institute of Science Technology (KAIST) Dept. of Civil and Environmental Engineering 3 Reinforcement Learning Policy

Korea Advanced Institute of Science Technology (KAIST)Dept. of Civil and Environmental Engineering

101

Support us in any formhttps://github.com/RLOpensource/tensorflow_RL/blob/master/model.py

https://github.com/RLOpensource/tensorflow_RL/blob/master/model.py

Ujava.org reinforcement-learning

Ujava.org reinforcement-learning

Data & Analytics
Introduction to Deep Reinforcement Learningwnzhang.net/teaching/cs420/slides/13-deep-rl.pdf · 2020-06-08 · Deep Reinforcement Learning •Deep Reinforcement Learning •leverages

Introduction to Deep Reinforcement Learningwnzhang.net/teaching/cs420/slides/13-deep-rl.pdf · 2020-06-08 · Deep Reinforcement Learning •Deep Reinforcement Learning •leverages

Documents
Reinforcement Learning Elementary Solution Methods

Reinforcement Learning Elementary Solution Methods

Documents
Safe and Efficient Off-Policy Reinforcement Learning

Safe and Efficient Off-Policy Reinforcement Learning

Software
Reinforcement Learning 2 - cs.uni-potsdam.de · Universität Potsdam Institut für Informatik Lehrstuhl Maschinelles Lernen Reinforcement Learning 2 Uwe Dick

Reinforcement Learning 2 - cs.uni-potsdam.de · Universität Potsdam Institut für Informatik Lehrstuhl Maschinelles Lernen Reinforcement Learning 2 Uwe Dick

Documents
REINFORCEMENT LEARNING Jesús Fernández Bes Noviembre 2012

REINFORCEMENT LEARNING Jesús Fernández Bes Noviembre 2012

Documents
Statistical Foundations of Reinforcement Learning: III

Statistical Foundations of Reinforcement Learning: III

Documents
Learning coordination strategies using reinforcement learning myriam z abramson , dissertation, 2003

Learning coordination strategies using reinforcement learning myriam z abramson , dissertation, 2003

Engineering
Deep Reinforcement Learning for Crazyhouse · Deep Reinforcement Learning for Crazyhouse MasterthesisbyJohannesCzech Dateofsubmission:December30,2019 1.Review:Prof.Dr.KristianKersting

Deep Reinforcement Learning for Crazyhouse · Deep Reinforcement Learning for Crazyhouse MasterthesisbyJohannesCzech Dateofsubmission:December30,2019 1.Review:Prof.Dr.KristianKersting

Documents
Reinforcement Learning 2

Reinforcement Learning 2

Documents
Učenje pojačavanjem Reinforcement Learning

Učenje pojačavanjem Reinforcement Learning

Documents
Reinforcement Learning Eligibility Traces

Reinforcement Learning Eligibility Traces

Documents
Optimizing Dialogue Management with Reinforcement Learning

Optimizing Dialogue Management with Reinforcement Learning

Documents
Chapter 13: Reinforcement Learning

Chapter 13: Reinforcement Learning

Documents
Reinforcement Learning Kurvendurchfahrtxubicomp/projekte/master08... · Thema Einsatz von Reinforcement Learning in der Modellfahrzeugnavigation Konkreter Anwendungsfall: Einführung

Reinforcement Learning Kurvendurchfahrtxubicomp/projekte/master08... · Thema Einsatz von Reinforcement Learning in der Modellfahrzeugnavigation Konkreter Anwendungsfall: Einführung

Documents
Reinforcement learning for e-marketing, presentation, 2012

Reinforcement learning for e-marketing, presentation, 2012

Technology
강화 학습 기초 Reinforcement Learning an introduction

강화 학습 기초 Reinforcement Learning an introduction

Technology
A Deep Reinforcement Learning Chatbot

A Deep Reinforcement Learning Chatbot

Data & Analytics
ujava.org workshop : Reinforcement Learning with Thompson Sampling

ujava.org workshop : Reinforcement Learning with Thompson Sampling

Data & Analytics
1 Einführung in das Reinforcement Learning DFG-Forschungsprojekt Einführung in das Reinforcement- Learning

1 Einführung in das Reinforcement Learning DFG-Forschungsprojekt Einführung in das Reinforcement- Learning

Documents
Reinforcement learning v0.5

Reinforcement learning v0.5

Data & Analytics
Reinforcement Learning Lecture Function Approximation

Reinforcement Learning Lecture Function Approximation

Documents
Meger ősítéses Tanulás = Reinforcement Learning (RL)

Meger ősítéses Tanulás = Reinforcement Learning (RL)

Documents
Deep Reinforcement Learning of Volume-Guided …openaccess.thecvf.com/content_CVPR_2019/papers/Han_Deep...Deep Reinforcement Learning of Volume-guided Progressive View Inpainting for

Deep Reinforcement Learning of Volume-Guided …openaccess.thecvf.com/content_CVPR_2019/papers/Han_Deep...Deep Reinforcement Learning of Volume-guided Progressive View Inpainting for

Documents
Reinforcement Learning - 4. Model-free reinforcement Learning

Reinforcement Learning - 4. Model-free reinforcement Learning

Documents
Reinforcement Learning

Reinforcement Learning

Documents
Reinforcement Learning Lecture Temporal Difference Learning

Reinforcement Learning Lecture Temporal Difference Learning

Documents
Hangzhou Deep Learning Meetup-Deep Reinforcement Learning

Hangzhou Deep Learning Meetup-Deep Reinforcement Learning

Technology
Hybrid Reinforcement Learning and Its Application to Biped ...papers.nips.cc/paper/1434-hybrid-reinforcement-learning...module reinforcement: r(t) 1-----10( decision of k Figure 3:

Hybrid Reinforcement Learning and Its Application to Biped ...papers.nips.cc/paper/1434-hybrid-reinforcement-learning...module reinforcement: r(t) 1-----10( decision of k Figure 3:

Documents
Reinforcement Learning: A Brief Introduction · 2012: Distributed deep learning (e.g., Google Brain) 2013: DQN for deep reinforcement learning 1997: LSTM-RNN 2015: Open source tools:

Reinforcement Learning: A Brief Introduction · 2012: Distributed deep learning (e.g., Google Brain) 2013: DQN for deep reinforcement learning 1997: LSTM-RNN 2015: Open source tools:

Documents