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. II Learning to Communicate
  2. 7. Bayesian Action Decoder
  3. 7.3 Method

7.3.4 Factorized Belief Updates

๋ชจ๋“  beliefs๋ฅผ ํ‘œํ˜„ํ•˜๋Š” ๊ฒƒ์€ ์–ด๋ ต์Šต๋‹ˆ๋‹ค. ์˜ˆ๋ฅผ ๋“ค๋ฉด, ์นด๋“œ ๊ฒŒ์ž„์—์„œ ๋ชจ๋“  agent๊ฐ€ ๋“ค๊ณ  ์žˆ์„ ์นด๋“œ ํŒจ์— ๋Œ€ํ•ด ๊ฐ€๋Šฅํ•œ ๊ฐœ์ˆ˜๋ฅผ ๋‚˜ํƒ€๋‚ด๋ ค๋ฉด exponentialํ•œ ๊ฒฝ์šฐ์˜ ์ˆ˜๋ฅผ ๊ฐ€์ง‘๋‹ˆ๋‹ค. ์ด๋Ÿฌํ•œ ์ƒํ™ฉ์„ ํ”ผํ•˜๊ธฐ ์œ„ํ•ด, belief state์— ๋Œ€ํ•ด ์ด์ „์— ฯ€^ \hat{\pi}ฯ€^๋ฅผ ๊ฐ€์ •ํ–ˆ ๊ฒƒ ์ฒ˜๋Ÿผ ๋‹ค์Œ๊ณผ ๊ฐ™์ด ๊ทผ์‚ฌํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

P(ftpriโˆฃfโ‰คtpub)โ‰ˆโˆiP(ftpri[i]โˆฃfโ‰คtpub)=Btfact P(f^{\mathrm{pri}}_t|f^{\mathrm{pub}}_{\leq t}) \approx \prod_i P(f^{\mathrm{pri}}_t[i]|f^{\mathrm{pub}}_{\leq t})= \mathcal{B}^{\mathrm{fact}}_{t}P(ftpriโ€‹โˆฃfโ‰คtpubโ€‹)โ‰ˆโˆiโ€‹P(ftpriโ€‹[i]โˆฃfโ‰คtpubโ€‹)=Btfactโ€‹

์ด์ œ ์ด factorized belief์— ๋Œ€ํ•ด superscription์„ ์—†์•ค ์ฑ„๋กœ ์‚ฌ์šฉํ•˜๊ฒ ์Šต๋‹ˆ๋‹ค. ๋‹ค์‹œ ์นด๋“œ ๊ฒŒ์ž„์˜ ์˜ˆ๋ฅผ ๋“ค๋ฉด, ๊ฐ ์š”์†Œ๋Š” ์นด๋“œ๋‹น ํ™•๋ฅ  ๋ถ„ํฌ๋ฅผ ๋‚˜ํƒ€๋‚ด๋Š”๋ฐ, ์ด๋•Œ ์†์— ๋“  ์นด๋“œ์™€ ํ”Œ๋ ˆ์ด์–ด๋“ค ์‚ฌ์ด์˜ ์นด๋“œ๊ฐ€ ๋…๋ฆฝ์ ์ด๋ผ๊ณ  ๊ฐ€์ •ํ•œ ๊ฒƒ๊ณผ ๊ฐ™์Šต๋‹ˆ๋‹ค. ์ด๋Ÿฌํ•œ ๊ฐ€์ •์€ ๋‹ค๋ฅธ ๋‹ค๋ฃจ๊ธฐ ์–ด๋ ค์šด ์ƒํ™ฉ์—์„œ๋„ ๋ฌธ์ œ๋ฅผ ๋‹ค๋ฃจ๊ธฐ ์‰ฝ๊ฒŒ ๊ทผ์‚ฌํ•˜๋Š” ๋ฐฉ๋ฒ•์ค‘์— ํ•˜๋‚˜์ž…๋‹ˆ๋‹ค.

public belief update๋ฅผ ์ด factorized representation๊ณผ ํ•จ๊ป˜ ์‚ฌ์šฉํ•˜๊ธฐ ์œ„ํ•ด, factorized likelihood term Lt[f[i]]\mathcal{L}_t[f[i]]Ltโ€‹[f[i]]์„ ๊ฐ private feature์— ์žฌ๊ท€์ ์œผ๋กœ ์ ์šฉํ•ฉ๋‹ˆ๋‹ค.

Lt[f[i]]=P(uโ‰คtaโˆฃf[i],Bโ‰คt,fโ‰คtpub,ฯ€^โ‰คt) \mathcal{L}_t[f[i]] = P (u^a_{\leq{t}}|f[i],\mathcal{B}_{\leq t}, f^{\mathrm{pub}}_{\leq t } , \hat{\pi}_{\leq t})Ltโ€‹[f[i]]=P(uโ‰คtaโ€‹โˆฃf[i],Bโ‰คtโ€‹,fโ‰คtpubโ€‹,ฯ€^โ‰คtโ€‹)

โ‰ˆLtโˆ’1[f[i]]โ‹…P(utaโˆฃf[i],Bt,ftpub,ฯ€^t) \approx \mathcal{L}_{t-1}[f[i]] \cdot P(u^a_t|f[i],\mathcal{B}_t,f^{\mathrm{pub}}_t,\hat{\pi}_t)โ‰ˆLtโˆ’1โ€‹[f[i]]โ‹…P(utaโ€‹โˆฃf[i],Btโ€‹,ftpubโ€‹,ฯ€^tโ€‹)

=Ltโˆ’1[f[i]]โ‹…EftโˆผBt[1(ft[i],f[i])1(ฯ€^(fta),uta)]EftโˆผBt[1(ft[i],f[i])]= L_{t-1}[f[i]] \cdot \frac{\mathbb{E}_{f_t \sim \mathcal{B}_t}[\bm{1}(f_t[i],f[i])\bm{1}(\hat{\pi}(f^a_t),u^a_t)]}{\mathbb{E}_{f_t\sim \mathcal{B}_t}[\bm{1}(f_t[i],f[i])]}=Ltโˆ’1โ€‹[f[i]]โ‹…Eftโ€‹โˆผBtโ€‹โ€‹[1(ftโ€‹[i],f[i])]Eftโ€‹โˆผBtโ€‹โ€‹[1(ftโ€‹[i],f[i])1(ฯ€^(ftaโ€‹),utaโ€‹)]โ€‹

๋งˆ์ง€๋ง‰ ํ…€์—์„œ ๋ณด๋‹ค์‹œํ”ผ expectation์œผ๋กœ ๋‚˜ํƒ€๋‚˜๋Š”๋ฐ ์ด๋Š” sampling์„ ํ†ตํ•ด์„œ ๊ณ„์‚ฐํ•  ์ˆ˜ ์žˆ๊ณ , sample ๊ฐœ์ˆ˜๊ฐ€ ๋Š˜์–ด๋‚ ์ˆ˜๋ก ์ •ํ™•ํ•ด์ง‘๋‹ˆ๋‹ค.

Previous7.3.3 Sampling Deterministic Partial PoliciesNext7.3.5 Self-Consistent Beliefs

Last updated 4 years ago

Was this helpful?