상세정보

CONTENTS
Preface and Acknowledgments
1 Introduction = 1
 1.1 The Need for Generalizing Explanation Structures = 1
 1.2 Overview of this Book = 4
  1.2.1 Chapter Summaries = 5
  1.2.2 Relevance to Research Areas Outside Machine Learning = 6
 1.3 Explanation-Based Learning = 7
  1.3.1 A Brief History = 8
  1.3.2 The Standard Method = 9
  1.3.3 Additional Research Issues = 11
2 Learning in Mathematically-Based Domains = 15
 2.1 The PHYSICS 101 System = 15
  2.1.1 The Learning Model = 16
  2.1.2 Terminology = 17
  2.1.3 Other Approaches to Learning in Mathematical Domains = 18
 2.2 Solving Problems = 19
  2.2.1 Initial Knowledge of the System = 22
  2.2.2 Schema-Based Problem Solving = 25
  2.2.3 Choosing the Initial Equation = 26
  2.2.4 Transforming an Expression into an Acceptable Form = 28
 2.3 Building Explanations = 35
  2.3.1 A Sample Problem = 35
  2.3.2 Verifying a Teacher's Solution = 37
  2.3.3 Explanining Solutions = 40
  2.3.4 Understanding Obstacles = 42
  2.3.5 Constructing the Cancellation Graph-Algorithmic Details = 44
 2.4 Generalizing Solutions = 51
  2.4.1 The Result of Standard Explanation - Based Learning = 51
  2.4.2 Using the Cancellation Graph to guide Generalization = 53
  2.4.3 Learning Special-Case Schemata = 63
  2.4.4 Performance Analysis = 68
 2.5 Summary = 70
3 A Domain-Independent Approach = 71
 3.1 The BAGGER System = 71
  3.1.1 Some Sample Learning Episodes = 72
  3.1.2 Situation Calculus = 75
  3.1.3 Sequential rules = 76
  3.1.4 Representing Sequential Knowledge = 77
 3.2 Generalizing = 78
  3.2.1 The BAGGER Generalization Algorithm = 78
  3.2.2 Problem Solving in BAGGER = 87
  3.2.3 Simplifying the Antecedents in Sequential Rules = 88
  3.2.4 Two Examples = 92
 3.3 Extending BAGGER = 101
  3.3.1 Algorithmic Details and Correctness Proof = 102
  3.3.2 The Circuit Implementation Domain Revisited = 106
  3.3.3 Learning from Multiple Examples = 108
  3.3.4 Problem Solving with Rules Acquired by BAGGER2 = 109
  3.3.5 Improving the Efficiency of the Rules BAGGER2 Learns = 109
  3.3.6 Learning about Wagons = 114
  3.3.7 Comparing BAGGER and BAGGER2 = 114
 3.4 Summary = 115
4 An Empirical Analysis of Explanation-Based Learning = 117
 4.1 Introduction = 117
 4.2 Experimental Methodology = 117
 4.3 Experiments = 121
  4.3.1 Comparison of the Two Training Strategies = 121
  4.3.2 Effect of Increased Problem Complexity = 127
  4.3.3 Operationality versus Generality = 129
  4.3.4 Time Spent Learning = 132
  4.3.5 Clearing Blocks = 132
  4.3.6 Rule Access Strategies = 134
  4.3.7 Estimating the Performance of the Non-Learning System = 140
  4.3.8 Empirical Study of BAGGER2 = 140
 4.4 Discussion = 144
5 Conclusion = 147
 5.1 Contributions = 147
 5.2 Relation to Other Work = 149
  5.2.1 Other Explanation-Based Approaches = 149
  5.2.2 Related Work in Similarity-Based Learning = 151
  5.2.3 Related Work in Automatic Programming = 151
 5.3 Some Open Research Issues = 152
  5.3.1 Deciding When to Learn = 152
  5.3.2 Improving What is Learned = 154
  5.3.3 Extending What can be Learned = 155
  5.3.4 Additional Issues = 157
 5.4 Final Summary = 158
Appendix A Additional PHYSICS 101 Examples = 159
 A.1 Overview = 159
 A.2 Learning about Energy Conservation = 159
 A.3 Learning about the Sum of Internal Forces = 165
 A.4 Using the New Force Law to Learn about Momentum = 168
 A.5 Attempting to Learn from a Two-Ball Collision = 170
Appendix B Additional BAGGER Examples = 173
 B.1 Overview = 173
 B.2 More Tower-Building Rules = 173
 B.3 Clearing an Object = 179
 B.4 Setting Table = 185
Appendix C BAGGER'S Initial Inference Rules = 191
 C.1 Notation = 191
 C.2 Rules = 191
Appendix D Statistics from Experiments = 199
 D.1 Description = 199
 D.2 Statistics = 199
References = 203