Contents
    Preface
    Series Forward
    Summary of Notation


  I. The Problem
    1. Introduction
      1.1 Reinforcement Learning
      1.2 Examples
      1.3 Elements of Reinforcement Learning
      1.4 An Extended Example: Tic-Tac-Toe
      1.5 Summary
      1.6 History of Reinforcement Learning
      1.7 Bibliographical Remarks
    2. Evaluative Feedback
      2.1 An -Armed Bandit Problem
      2.2 Action-Value Methods
      2.3 Softmax Action Selection
      2.4 Evaluation Versus Instruction
      2.5 Incremental Implementation
      2.6 Tracking a Nonstationary Problem
      2.7 Optimistic Initial Values
      2.8 Reinforcement Comparison
      2.9 Pursuit Methods
      2.10 Associative Search
      2.11 Conclusions
      2.12 Bibliographical and Historical Remarks
    3. The Reinforcement Learning Problem
      3.1 The Agent-Environment Interface
      3.2 Goals and Rewards
      3.3 Returns
      3.4 Unified Notation for Episodic and Continuing Tasks
      3.5 The Markov Property
      3.6 Markov Decision Processes
      3.7 Value Functions
      3.8 Optimal Value Functions
      3.9 Optimality and Approximation
      3.10 Summary
      3.11 Bibliographical and Historical Remarks


  II. Elementary Solution Methods
    4. Dynamic Programming
      4.1 Policy Evaluation
      4.2 Policy Improvement
      4.3 Policy Iteration
      4.4 Value Iteration
      4.5 Asynchronous Dynamic Programming
      4.6 Generalized Policy Iteration
      4.7 Efficiency of Dynamic Programming
      4.8 Summary
      4.9 Bibliographical and Historical Remarks
    5. Monte Carlo Methods
      5.1 Monte Carlo Policy Evaluation
      5.2 Monte Carlo Estimation of Action Values
      5.3 Monte Carlo Control
      5.4 On-Policy Monte Carlo Control
      5.5 Evaluating One Policy While Following Another
      5.6 Off-Policy Monte Carlo Control
      5.7 Incremental Implementation
      5.8 Summary
      5.9 Bibliographical and Historical Remarks
    6. Temporal-Difference Learning
      6.1 TD Prediction
      6.2 Advantages of TD Prediction Methods
      6.3 Optimality of TD(0)
      6.4 Sarsa: On-Policy TD Control
      6.5 Q-Learning: Off-Policy TD Control
      6.6 Actor-Critic Methods
      6.7 R-Learning for Undiscounted Continuing Tasks
      6.8 Games, Afterstates, and Other Special Cases
      6.9 Summary
      6.10 Bibliographical and Historical Remarks


  III. A Unified View
    7. Eligibility Traces
      7.1 -Step TD Prediction
      7.2 The Forward View of TD()
      7.3 The Backward View of TD()
      7.4 Equivalence of Forward and Backward Views
      7.5 Sarsa()
      7.6 Q()
      7.7 Eligibility Traces for Actor-Critic Methods
      7.8 Replacing Traces
      7.9 Implementation Issues
      7.10 Variable  
      7.11 Conclusions
      7.12 Bibliographical and Historical Remarks
    8. Generalization and Function Approximation
      8.1 Value Prediction with Function Approximation
      8.2 Gradient-Descent Methods
      8.3 Linear Methods
        8.3.1 Coarse Coding
        8.3.2 Tile Coding
        8.3.3 Radial Basis Functions
        8.3.4 Kanerva Coding
      8.4 Control with Function Approximation
      8.5 Off-Policy Bootstrapping
      8.6 Should We Bootstrap?
      8.7 Summary
      8.8 Bibliographical and Historical Remarks
    9. Planning and Learning
      9.1 Models and Planning
      9.2 Integrating Planning, Acting, and Learning
      9.3 When the Model Is Wrong
      9.4 Prioritized Sweeping
      9.5 Full vs. Sample Backups
      9.6 Trajectory Sampling
      9.7 Heuristic Search
      9.8 Summary
      9.9 Bibliographical and Historical Remarks
    10. Dimensions of Reinforcement Learning
      10.1 The Unified View
      10.2 Other Frontier Dimensions
    11. Case Studies
      11.1 TD-Gammon
      11.2 Samuel's Checkers Player
      11.3 The Acrobot
      11.4 Elevator Dispatching
      11.5 Dynamic Channel Allocation
      11.6 Job-Shop Scheduling

想钱想疯了把！有人买吗？网上都有！

  想钱想疯了的楼主