๐Ÿ˜‡
Deep Multi-Agent Reinforcement Learning
  • Deep Multi-Agent Reinforcement Learning
  • Abstract & Contents
    • Abstract
  • 1. Introduction
    • 1. INTRODUCTION
      • 1.1 The Industrial Revolution, Cognition, and Computers
      • 1.2 Deep Multi-Agent Reinforcement-Learning
      • 1.3 Overall Structure
  • 2. Background
    • 2. BACKGROUND
      • 2.1 Reinforcement Learning
      • 2.2 Multi-Agent Settings
      • 2.3 Centralized vs Decentralized Control
      • 2.4 Cooperative, Zero-sum, and General-Sum
      • 2.5 Partial Observability
      • 2.6 Centralized Training, Decentralized Execution
      • 2.7 Value Functions
      • 2.8 Nash Equilibria
      • 2.9 Deep Learning for MARL
      • 2.10 Q-Learning and DQN
      • 2.11 Reinforce and Actor-Critic
  • I Learning to Collaborate
    • 3. Counterfactual Multi-Agent Policy Gradients
      • 3.1 Introduction
      • 3.2 Related Work
      • 3.3 Multi-Agent StarCraft Micromanagement
      • 3.4 Methods
        • 3.4.1 Independent Actor-Critic
        • 3.4.2 Counterfactual Multi-Agent Policy Gradients
        • 3.4.2.1 baseline lemma
        • 3.4.2.2 COMA Algorithm
      • 3.5 Results
      • 3.6 Conclusions & Future Work
    • 4 Multi-Agent Common Knowledge Reinforcement Learning
      • 4.1 Introduction
      • 4.2 Related Work
      • 4.3 Dec-POMDP and Features
      • 4.4 Common Knowledge
      • 4.5 Multi-Agent Common Knowledge Reinforcement Learning
      • 4.6 Pairwise MACKRL
      • 4.7 Experiments and Results
      • 4.8 Conclusion & Future Work
    • 5 Stabilizing Experience Replay
      • 5.1 Introduction
      • 5.2 Related Work
      • 5.3 Methods
        • 5.3.1 Multi-Agent Importance Sampling
        • 5.3.2 Multi-Agent Fingerprints
      • 5.4 Experiments
        • 5.4.1 Architecture
      • 5.5 Results
        • 5.5.1 Importance Sampling
        • 5.5.2 Fingerprints
        • 5.5.3 Informative Trajectories
      • 5.6 Conclusion & Future Work
  • II Learning to Communicate
    • 6. Learning to Communicate with Deep Multi-Agent ReinforcementLearning
      • 6.1 Introduction
      • 6.2 Related Work
      • 6.3 Setting
      • 6.4 Methods
        • 6.4.1 Reinforced Inter-Agent Learning
        • 6.4.2 Differentiable Inter-Agent Learning
      • 6.5 DIAL Details
      • 6.6 Experiments
        • 6.6.1 Model Architecture
        • 6.6.2 Switch Riddle
        • 6.6.3 MNIST Games
        • 6.6.4 Effect of Channel Noise
      • 6.7 Conclusion & Future Work
    • 7. Bayesian Action Decoder
      • 7.1 Introduction
      • 7.2 Setting
      • 7.3 Method
        • 7.3.1 Public belief
        • 7.3.2 Public Belief MDP
        • 7.3.3 Sampling Deterministic Partial Policies
        • 7.3.4 Factorized Belief Updates
        • 7.3.5 Self-Consistent Beliefs
      • 7.4 Experiments and Results
        • 7.4.1 Matrix Game
        • 7.4.2 Hanabi
        • 7.4.3 Observations and Actions
        • 7.4.4 Beliefs in Hanabi
        • 7.4.5 Architecture Details for Baselines and Method
        • 7.4.6 Hyperparamters
        • 7.4.7 Results on Hanabi
      • 7.5 Related Work
        • 7.5.1 Learning to Communicate
        • 7.5.2 Research on Hanabi
        • 7.5.3 Belief State Methods
      • 7.6 Conclusion & Future Work
  • III Learning to Reciprocate
    • 8. Learning with Opponent-Learning Awareness
      • 8.1 Introduction
      • 8.2 Related Work
      • 8.3 Methods
        • 8.3.1 Naive Learner
        • 8.3.2 Learning with Opponent Learning Awareness
        • 8.3.3. Learning via Policy gradient
        • 8.3.4 LOLA with Opponent modeling
        • 8.3.5 Higher-Order LOLA
      • 8.4 Experimental Setup
        • 8.4.1 Iterated Games
        • 8.4.2 Coin Game
        • 8.4.3 Training Details
      • 8.5 Results
        • 8.5.1 Iterated Games
        • 8.5.2 Coin Game
        • 8.5.3 Exploitability of LOLA
      • 8.6 Conclusion & Future Work
    • 9. DiCE: The Infinitely Differentiable Monte Carlo Estimator
      • 9.1 Introduction
      • 9.2 Background
        • 9.2.1 Stochastic Computation Graphs
        • 9.2.2 Surrogate Losses
      • 9.3 Higher Order Gradients
        • 9.3.1 Higher Order Gradient Estimators
        • 9.3.2 Higher Order Surrogate Losses
        • 9.3.3. Simple Failing Example
      • 9.4 Correct Gradient Estimators with DiCE
        • 9.4.1 Implement of DiCE
        • 9.4.2 Casuality
        • 9.4.3 First Order Variance Reduction
        • 9.4.4 Hessian-Vector Product
      • 9.5 Case Studies
        • 9.5.1 Empirical Verification
        • 9.5.2 DiCE For multi-agent RL
      • 9.6 Related Work
      • 9.7 Conclusion & Future Work
  • Reference
    • Reference
  • After
    • ๋ณด์ถฉ
    • ์—ญ์ž ํ›„๊ธฐ
Powered by GitBook
On this page

Was this helpful?

  1. III Learning to Reciprocate
  2. 9. DiCE: The Infinitely Differentiable Monte Carlo Estimator
  3. 9.4 Correct Gradient Estimators with DiCE

9.4.1 Implement of DiCE

Previous9.4 Correct Gradient Estimators with DiCENext9.4.2 Casuality

Last updated 4 years ago

Was this helpful?

DiCE์˜ ์ฒ˜์Œ๋ถ€ํ„ฐ ์‹ค์šฉ์ ์ธ ๋…ผ๋ฌธ์ž„์„ ๊ฐ•์กฐํ–ˆ๋Š”๋ฐ, ์ด๋Š” ๋‹จ์ˆœํ•œ ๊ตฌํ˜„๋ฐฉ๋ฒ•์— ์žˆ์Šต๋‹ˆ๋‹ค. MagicBox๋Š” ๋‘๊ฐ€์ง€ ํŠน์„ฑ์„ ๋งŒ์กฑํ•˜๋ฉด ๋์—ˆ๋Š”๋ฐ, ์ด๋Š” ๋‹ค์Œ๊ณผ ๊ฐ™์ด ์ •์˜ํ•จ์œผ๋กœ์จ ๋‘๊ฐ€์ง€ ์„ฑ์งˆ์„ ๋‹ค ๊ฐ€์ ธ๊ฐˆ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ์•„๋ž˜์„  ์œ„๋ฅผ ํ™•์ธํ•˜๋„๋ก ํ•ฉ๋‹ˆ๋‹ค.

โ–ก(W)=expโก(ฯ„โˆ’โŠฅ(ฯ„)) \square (\mathcal{W}) = \exp(\tau - \bot(\tau))โ–ก(W)=exp(ฯ„โˆ’โŠฅ(ฯ„))

ฯ„=โˆ‘wโˆˆWlogโก(p(w;ฮธ)) \tau = \sum_{w \in \mathcal{W}} \log(p(w;\theta))ฯ„=โˆ‘wโˆˆWโ€‹log(p(w;ฮธ))

โŠฅ \bot โŠฅ์€ โˆ‡xโŠฅ(x)=0 \nabla_x \bot(x) = 0 โˆ‡xโ€‹โŠฅ(x)=0์ด ๋˜๋„๋ก ํ•˜๋Š” gradient๋ฅผ ์•ˆํ๋ฅด๋„๋ก ํ•˜๋Š” operator๋กœ pytorch์˜ detach๊ฐ™์€ ์—ญํ• ์„ ํ•ฉ๋‹ˆ๋‹ค. โŠฅ(x)โ†’x\bot(x) \rightarrow x โŠฅ(x)โ†’x์ด๋ฏ€๋กœ, โ–ก(W)โ†’1\square (\mathcal{W}) \rightarrow 1โ–ก(W)โ†’1์ž„์ด ์ž๋ช…ํ•ฉ๋‹ˆ๋‹ค. ์ด๋กœ์จ ์ฒซ๋ฒˆ์งธ ์„ฑ์งˆ์ด ์ฆ๋ช…๋˜์—ˆ์Šต๋‹ˆ๋‹ค. ๋ฐ”๋กœ ๋‘๋ฒˆ์งธ ์„ฑ์งˆ์„ ์ฆ๋ช…ํ•˜๋ฉด ๋‹ค์Œ๊ณผ ๊ฐ™์Šต๋‹ˆ๋‹ค.

โˆ‡ฮธโ–ก(W)=โˆ‡ฮธexpโก(ฯ„โˆ’โŠฅ(ฯ„)) \nabla_\theta \square (\mathcal{W}) = \nabla_\theta\exp(\tau-\bot(\tau))โˆ‡ฮธโ€‹โ–ก(W)=โˆ‡ฮธโ€‹exp(ฯ„โˆ’โŠฅ(ฯ„))

=expโก(ฯ„โˆ’โŠฅ(ฯ„))โˆ‡ฮธ(ฯ„โˆ’โŠฅ(ฯ„))= \exp(\tau-\bot(\tau))\nabla_\theta(\tau-\bot(\tau))=exp(ฯ„โˆ’โŠฅ(ฯ„))โˆ‡ฮธโ€‹(ฯ„โˆ’โŠฅ(ฯ„))

=โ–ก(W)(โˆ‡ฮธฯ„โˆ’0)=\square(\mathcal{W})(\nabla_\theta\tau-0)=โ–ก(W)(โˆ‡ฮธโ€‹ฯ„โˆ’0)

=โ–ก(W)โˆ‘wโˆˆWโˆ‡ฮธlogโก(p(w;ฮธ))=\square(\mathcal{W})\sum_{w \in \mathcal{W}}\nabla_\theta \log(p(w;\theta))=โ–ก(W)โˆ‘wโˆˆWโ€‹โˆ‡ฮธโ€‹log(p(w;ฮธ))

๊ทธ๋ฆฌ๊ณ  magicbox operator๋ฅผ ๊ตฌํ˜„ํ•˜๊ฒŒ๋˜๋ฉด, ์ฃผ๋กœ objective์™€ ๋ฐ”๋กœ ์—ฐ๊ด€์ง€์–ด ๊ตฌํ˜„ํ•˜๋Š”๊ฒŒ ๊ฐ€์žฅ ๊ฐ„๋‹จํ•œ๋ฐ, ์ผ๋ฐ˜์ ์ธ RL์—์„  J=E[โˆ‘rt]J = \mathbb{E}[\sum r_t]J=E[โˆ‘rtโ€‹]๋กœ ๋‚˜ํƒ€๋‚ผ ๋•Œ, DiCE์˜ objective๋Š” Jโ–ก=โˆ‘tโ–ก({atโ€ฒ,tโ€ฒโ‰คt})rt J_\square= \sum_t \square(\{a_{t'},t'\leq t\})r_tJโ–กโ€‹=โˆ‘tโ€‹โ–ก({atโ€ฒโ€‹,tโ€ฒโ‰คt})rtโ€‹๋กœ ๋‚˜ํƒ€๋‚ด์•ผํ•ฉ๋‹ˆ๋‹ค. (์ด๋Š” ์ด์ „์˜ action์— ๋”ฐ๋ผ reward์— stochasticํ•˜๊ฒŒ ์˜ํ–ฅ์„ ์ฃผ๊ธฐ ๋•Œ๋ฌธ์ž…๋‹ˆ๋‹ค.