TOC Front Cover 2 Machine Learning 3 Copyright Page 4 Table of Contents 5 Preface 9 Program Committee 10 ML 92 Informal Workshop Themes and Coordinators 10 Chapter 1. Generalizing from Case Studies: A Case Study 11 Abstract 11 1 PROBLEM AND OBJECTIVES 11 2 GENERALIZING CASE STUDIES 12 3 AN APPLICATION 13 4 LIMITATIONS 19 References 20 5 CONCLUSION 20 Front Cover 2 Machine Learning 3 Copyright Page 4 Table of Contents 5 Preface 9 Program Committee 10 ML 92 Informal Workshop Themes and Coordinators 10 Chapter 1. Generalizing from Case Studies: A Case Study 11 Abstract 11 1 PROBLEM AND OBJECTIVES 11 2 GENERALIZING CASE STUDIES 12 3 AN APPLICATION 13 4 LIMITATIONS 19 References 20 5 CONCLUSION 20 Acknowledgements 20 Chapter 2. On Learning More Concepts 21 1 INTRODUCTION 21 Abstract 21 4 THE MULTI-BALLS LEARNING ALGORITHM 23 3 UPPER BOUND ON COVERAGE 23 5 THE LARGE-BALL LEARNING ALGORITHM 26 6 COVERAGE OF CURRENT LEARNING ALGORITHMS 27 7 DISCUSSION 28 Acknowledgements 28 References 28 Chapter 3. The Principal Axes Method for Constructive Induction 30 1 INTRODUCTION 30 Abstract 30 2 LEARNING PRINCIPAL AXES 31 4 GENERATING SIMILARITY MATRIX 32 3 DISTANCE METRIC 32 5 DESCRIPTION SPACE TRANSFORMATION 33 7 SUMMARY 35 6 EMPIRICAL EVALUATION 35 Acknowledgments 36 References 37 Chapter 4. Learning by Incomplete Explanations of Failures in Recursive Domains 40 1 Introduction 40 2 Means-ends analysis search in recursive domains 40 Abstract 40 3 Problem solving and learning in FS2 41 4 Experimental results 43 5 Related work 44 6 Conclusions and future work 45 References 45 Chapter 5. Eliminating Redundancy in Explanation-Based Learning 47 Abstract 47 1 INTRODUCTION 47 3 EXAMPLE-GUIDEDUNFOLDING 48 2 PRELIMINARIES 48 5 RELATED WORK 50 4 EXPERIMENTAL RESULTS 50 6 CONCLUDING REMARKS 51 References 51 Chapter 6. Trading off Consistency and Efficiency in Version-Space Induction 53 Abstract 53 1 INTRODUCTION 53 2 LEARNING WITH VARIABLE-FACTORED CONJUNCTIVE CONCEPT LANGUAGES 54 3 THE FCE LEARNING ALGORITHM 54 4 UTILITY 56 Acknowledgements 58 5 RELATION TO INDUCTIVE LANGUAGE SHIFT 58 References 58 6 CONCLUSION 58 Chapter 7. Peepholing: choosing attributes efficiently for megainduction 59 Abstract 59 1 INTRODUCTION AND MOTIVATION 59 2 PEEPHOLING 60 3 SHORTLISTING 60 4 BLINKERING 62 5 EMPIRICAL EVALUATION 63 References 64 6 CONCLUSIONS 64 Acknowledgements 64 Chapter 8. Improving Path Planning with Learning 65 Abstract 65 1 INTRODUCTION 65 2 ALGORITHM 65 3 GENERAL ANALYSIS 67 4 SPECIFIC CASE ANALYSIS 69 5 COMPUTATIONAL EXPERIENCE 70 Acknowledgements 71 6 FUTURE WORK 71 References 71 7 CONCLUSION 71 CHAPTER 9. THE RIGHT REPRESENTATION FOR DISCOVERY: FINDING THE CONSERVATION OF MOMENTUM 72 Abstract 72 1 INTRODUCTION 72 3 CONVENTIONAL MATHEMATICAL APPROACH 73 2 CONSERVATION OF MOMENTUM 73 4 THE DIAGRAMMATIC APPROACH 76 5 DISCUSSION 80 References 81 6 CONCLUSIONS 81 Acknowledgements 81 Chapter 10. Learning to Predict in Uncertain Continuous Tasks 82 1 Introduction 82 Abstract 82 3 Manipulation Tasks 83 2 Assumptions 83 6 Funnels 84 4 Noise and Uncertainty 84 5 Generalization 84 7 Learning Funnels 85 8 Experiments 88 9 Assumptions Revisited 90 References 91 Acknowledgements 91 Chapter 11. Lazy Partial Evaluation: An Integration of Explanation-Based Generalisation and Partial Evaluation 92 1 Introduction 92 Abstract 92 2 Lazy Partial Evaluation 93 3 Application to Constraint Satisfaction 95 Acknowledgements and Availability 99 4 Discussion 99 References 99 5 Conclusion 99 Chapter 12. A Teaching Method for Reinforcement Learning 102 Abstract 102 1 INTRODUCTION 102 3 INTEGRATING THE METHOD 103 2 TEACHING METHOD 103 4 EXPERIMENT ONE: CART-POLE AND ACE/ASE 104 5 EXPERIMENT TWO: RACE TRACK AND Q-LEARNING 107 6 CONCLUSIONS 110 References 111 Acknowledgments 111 Chapter 13. Compiling Prior Knowledge Into an Explicit Bias 112 1 PROBLEMS FACING THEORY-GUIDED LEARNING 112 Abstract 112 3 ANTECEDENT DESCRIPTION GRAMMARS 114 2 EXPLICITLY BIASED LEARNING 114 4 EXPERIMENTAL RESULTS 117 5 CONCLUSIONS 118 Acknowledgements 119 A GRENDEL's LEARNING ALGORITHM 119 References 120 Chapter 14. Spatial analogy and subsumption 121 2 Spatio–analogical inference 121 1 Introduction 121 3 Subsumption and generalization 122 4 Retrieval and indexing 123 5 Results 125 6 Discussion 125 References 126 Acknowledgements 126 Chapter 15. Learning to Satisfy Conjunctive Goals 127 Abstract 127 1 Motivation 127 2 Opportunism and Learning 128 3 An example 129 4 When are plans for conjunctive goals necessary? 130 5 Reasons for saving plans for conjunctive goals 130 6 Stability and Enforcement 131 7 Stability and evaluation 131 8 Conclusion 132 Acknowledgements 132 References 132 Chapter 16. Multistrategy Learning with Introspective Meta-Explanations 133 Abstract 133 1 INTRODUCTION 133 2 REPRESENTATION OF INTROSPECTIVE META-XPS 134 Acknowledgements 137 References 137 3 DISCUSSION 137 Chapter 17. An Asymptotic Analysis of Speedup Learning 139 Abstract 139 1 Introduction 139 2 Meta-Level Problem Solvers 139 3 Polynomial-time Problem Solving 142 4 Macro Problem Solvers 142 5 Conclusion 144 Acknowledgments 145 References 145 Chapter 18. Why EBL Produces Overly-Specific Knowledge: A Critique of the PRODIGY Approaches 147 Abstract 147 1 Motivation 147 2 The Problem of Logical Minimization 148 3 Avoiding Overly-Specific Conditions 149 4 Related Work 152 5 Conclusion 153 References 153 Chapter 19. Automatic Feature Generation for Problem Solving Systems 154 Abstract 154 1 INTRODUCTION 154 2 PROBLEM SOLVING, EVALUATION FUNCTIONS AND FEATURES 155 3 A THEORY OF FEATURE GENERATION 155 4 THE ZENITH SYSTEM 156 5 DOMAINS AND RESULTS 160 Acknowledgements 163 References 163 6 CONCLUSION 163 Chapter 20. Towards Inductive Generalisation in Higher Order Logic 164 Abstract 164 1 Introduction 164 2 Motivation 165 3 Mλ: a restricted higher order language 166 4 Mλ normal and nonredundant terms 166 5 Implementation 167 6 Applications 168 7 Conclusion and future research directions 169 Acknowledgements 170 A Syntax of λ-calculus terms 170 B Extension to δ0 conversion 171 References 171 Chapter 21. Ordering Effects in Clustering 173 Abstract 173 1 INTRODUCTION 173 2 A REVIEW OF COBWEB 173 3 ORDERING EFFECTS 174 4 CONTROL STRATEGIES 175 5 ORDER INDEPENDENCE 176 6 CONCLUDING REMARKS 178 References 178 Chapter 22. Learning Structured Concepts Using Genetic Algorithms 179 Abstract 179 1 INTRODUCTION 179 2 WHY A GENETIC ALGORITHM? 180 3 GA-SMART OVERVIEW 180 4 MAPPING FORMULAE TO BIT STRINGS 181 5 GENETIC ALGORITHM DETAILS 182 6. EXPERIMENTATION WITH STANDARD TEST CASES 184 7 DISCUSSION 187 Acknowledgements 187 References 187 Chapter 23. An Analysis of Learning to Plan as a Search Problem 189 1 INTRODUCTION 189 2 LEARNING AS SEARCH 189 Abstract 189 3 FRAMEWORK OF SIMPLIFICATIONS 190 4 APPLYING THE FRAMEWORK 193 References 197 5 CONCLUSIONS 197 Acknowledgements 197 Chapter 24. An Approach to Anytime Learning 199 Abstract 199 1 INTRODUCTION 199 2 AN ARCHITECTURE FOR ANYTIME LEARNING 200 3 A CASE STUDY 201 4 SUMMARY 204 References 205 Chapter 25. Artificial Universes - Towards a Systematic Approach to Evaluating Algorithms which Learn from Examples 206 1. INTRODUCTION 206 2. MODELLING NOISE IN THE DOMAIN RULES 206 Abstract 206 3. A UNIVERSE: A PROBABILISTIC MODEL OF A DOMAIN 207 4. CONSTRUCTING A UNIVERSE - A WORKED EXAMPLE 208 5. INFORMATION IN THE UNIVERSE -EVALUATING INDUCED RULES 208 6 EXPERIMENTS USING ID3 ON GENERATED EXAMPLES 212 7 CONCLUSION AND FUTURE WORK 214 References 214 Chapter 26. Average Case Analysis of Learning k-CNF Concepts 216 2 AN AVERAGE CASE MODEL OF k-CNF 216 Abstract 216 1 INTRODUCTION 216 3 IMPLICATIONS OF THE MODEL 219 5 CONCLUSION 220 4 VIOLATING ASSUMPTIONS 220 References 221 Acknowledgements 221 Chapter 27. The MENTLE Approach to Learning Heuristics for the Control of Logic Programs 222 1 INTRODUCTION 222 2 HEURISTICS FOR GOAL SELECTION 222 Abstract 222 3 MENTLE LEARNING ALGORITHM 223 4 CONTRADICTION 226 5 RESULTS FROM MENTLE 226 6 FURTHER DEVELOPMENT 227 References 227 Chapter 28. Fuzzy Substructure Discovery 228 Abstract 228 1 INTRODUCTION 228 2 SUBSTRUCTURE DISCOVERY 228 3 FUZZY GRAPH MATCH 230 4 FUZZY SUBSTRUCTURE DISCOVERY 231 5 EXAMPLES 232 6 CONCLUSIONS 233 References 233 Chapter 29. Efficient Classification of Massive, Unsegmented Datastreams 234 Abstract 234 1 INTRODUCTION 234 2 EXISTING ML APPROACHES FOR FINDING MOTIFS IN DATASTREAMS 235 3 OUR APPROACH 235 4. TESTING THE CLUSTERER: THE PROTEIN EVOLUTION SIMULATOR 239 5. APPLICATION TO REAL DATA 239 References 241 6. CONCLUSIONS 241 Chapter 30. Induction of One-Level Decision Trees 243 1 INTRODUCTION 243 2 ANALYSIS OF THE ONE-LEVEL ALGORITHM 243 Abstract 243 3 BEHAVIOR OF THE ONE-LEVEL ALGORITHM 246 4 DISCUSSION 249 References 250 Acknowledgements 250 Chapter 31. Combining Competition and Cooperation in Supervised Inductive Learning 251 1 Preliminaries 251 Abstract 251 2 DESIGN 252 3 EXPERIMENTS 255 4 SUMMARY 257 References 257 Chapter 32. A Practical Approach to Feature Selection 259 Abstract 259 Show more