Machine Learning Proceedings 1989 (eBook)

Alberto Maria Segre (Herausgeber)

eBook Download: PDF

2014 | 1. Auflage
510 Seiten
Elsevier Science (Verlag)
978-1-4832-9740-8 (ISBN)

Infos

Machine Learning Proceedings 1989

Inhalt

Front Cover 1
Proceedings of the Sixth International Workshop on Machine Learning 2
Copyright Page 3
Table of Contents 4
PREFACE 10
Part 1: Combining Empirical and Explanation-Based Learning 12
Chapter 
13 
The Need for Unified Theories of Learning 13
Learning from One Instance and Many Instances 13
Learning With and Without Search 13
Learning With and Without Domain Knowledge 14
Justified and Unjustified Learning 14
Accuracy and Efficiency in Machine Learning 15
CHAPTER 
16 
ABSTRACT 16
INTRODUCTION 16
THE IOU APPROACH 16
AN INITIAL IOU ALGORITHM 17
IOU VERSUS PURE SBL AND IOE 18
CONCLUSIONS AND FUTURE RESEARCH 18
CHAPTER 
19 
INDUCTION-BASED AND EXPLANATION-BASED LEARNING 19
OPEN PROBLEMS 19
CONCEPTUAL CLUSTERING OF EXPLANATIONS 20
CONCLUDING REMARKS 21
References 21
Chapter 
22 
1 Introduction 22
2 A new integration framework: generalized explanations 22
3 An application example 23
References 24
CHAPTER 
25 
ABSTRACT 25
INTRODUCTION 25
DISCIPLE AS AN EXPERT SYSTEM 25
THE LEARNING PROBLEM 25
LEARNING IN A COMPLETE THEORY DOMAIN 26
LEARNING IN A WEAK THEORY DOMAIN 26
CONCLUSIONS 27
References 27
CHAPTER 6. A DESCRIPTION OF PREFERENCE CRITERION IN CONSTRUCTIVE LEARNING: A Discussion of Basic Issues 28
1. INTRODUCTION 28
2. CONSTRUCTIVE LEARNING 28
3. INDIVIDUAL CRITERIA AND THEIR RELATIONSHIPS 29
Acknowledgements 30
Reference 30
CHAPTER 7. COMBINING CASE-BASED REASONING, EXPLANATION-BASED LEARNING, AND LEARNING FROM INSTRUCTION 31
ABSTRACT 31
INTRODUCTION 31
INFERRING IN STRUCTOR'S GOAL 32
INFERRING PLACE IN CURRENT DIAGNOSIS 32
ADJUSTING THE SALIENCE OF FEATURES 33
CAUSAL EXPLANATION OF ACTIONS 33
CONCLUSION 33
References 33
CHAPTER 8. DEDUCTION IN TOP-DOWN INDUCTIVE LEARNING 34
References 36
CHAPTER 9. ONE-SIDED ALGORITHMS FOR INTEGRATING EMPIRICAL AND EXPLANATION-BASED LEARNING 37
A FRAMEWORK FOR INTEGRATED LEARNING 37
PERFORMANCE AND FOUNDATIONAL EXAMPLES 37
THE IOSC and 
38 
CONCLUSION 39
References 39
CHAPTER 10. COMBINING EMPIRICAL AND ANALYTICAL LEARNING WITH VERSION SPACES 40
ABSTRACT 40
INTRODUCTION 40
USING INCREMENTAL VERSION-SPACE MERGING ON THE RESULTS OF EBG 41
PERSPECTIVES 42
RELATED WORK 43
SUMMARY 44
References 44
CHAPTER 11. FINDING NEW RULES FOR INCOMPLETE THEORIES: EXPLICIT BIASES FOR INDUCTION WITH CONTEXTUAL INFORMATION 45
INTRODUCTION 45
HEURISTICS EXPLOITING CONTEXTUAL INFORMATION AS A STRONG INDUCTIVE BIAS 45
EMPIRICAL SELECTION OF BIASES 46
CONCLUSION 46
Acknowledgments 47
REFERENCES 47
CHAPTER 12. LEARNING FROM PLAUSIBLE EXPLANATIONS 48
INTRODUCTION 48
THE LEARNING METHOD 48
CONCLUSION 50
References 50
CHAPTER 13. AUGMENTING DOMAIN THEORY FOR EXPLANATION-BASED GENERALISATION 51
INTRODUCTION 51
AUGMENTING THE DOMAIN THEORY 51
EXPERIMENTAL ANALYSIS 52
PROBLEMS AND FUTURE RESEARCH 53
CONCLUSION 53
References 53
Chapter 
54 
Introduction 54
An Example: The Mob System 54
Conclusion 56
References 56
CHAPTER 15. REDUCING SEARCH AND LEARNING GOAL PREFERENCES 57
INTRODUCTION 57
DEPTHFIRST SEARCH 57
FRAMEWORK FOR A SEARCH CONTROL HEURISTIC 58
A SEARCH ALGORITHM - DEFINITION AND RESULTS 58
References 59
Chapter 
60 
Summary 62
References 62
CHAPTER 17. A RETRIEVAL MODEL USING FEATURE SELECTION 63
ABSTRACT 63
A DISTRIBUTED MODEL OF RETRIEVAL 63
CONTROLUNG RETRIEVAL WITH GOALS 64
EXTENSIONS TO LEARNING 65
REFERENCES 65
CHAPTER 18. IMPROVING DECISION-MAKING ON THE BASIS OF EXPERIENCE 66
References 68
CHAPTER 19. EXPLANATION-BASED ACCELERATION OF SIMILARITY-BASED LEARNING 69
ABSTRACT 69
INTRODUCTION 69
REVERSIBLE INTERPRETER 69
EXPLANATION-BASED ACCELERATION 70
CONCLUSION 71
References 71
Chapter 
72 
An Introduction to KA Planning 72
IVY: A KA Planner in Lung Tumor Pathology 73
KA Planning for Scientific Discovery 74
Conclusion 76
References 76
Chapter 
77 
ABSTRACT 77
INTRODUCTION 77
A CONNECTIONIST KNOWLEDGE REPRESENTATION SYSTEM 78
CONCLUSION 79
References 79
CHAPTER 22. INTEGRATING LEARNING IN A NEURAL NETWORK 80
ABSTRACT 80
INTRODUCTION 80
ARCHITECTURE AND INFERENCE 80
SBL 81
EBL 81
RESULTS 82
References 82
Chapter 
83 
Acknowledgements 85
References 85
Chapter 
86 
1 Introduction 86
2 Definitions 86
3 Algorithm 87
4 Example 87
5 Conclusion 88
References 88
Chapter 
89 
Introduction 89
Representation and Performance 89
Learning Efficient and Accurate Domain Theories 90
Current Status 91
References 91
CHAPTER 26. ERROR CORRECTION IN CONSTRUCTIVE INDUCTION 92
ABSTRACT 92
1. INTRODUCTION 92
2. INDUCTION IN AN ABSTRACTION SPACE 92
3. CORRECTION OF ATTRIBUTE NOISE 93
References 94
CHAPTER 27. IMPROVING EXPLANATION-BASED INDEXING WITH EMPIRICAL LEARNING 95
ABSTRACT 95
INTRODUCTION 95
OVER-GENERLIZATION IN EBI 95
USING EMPIRICAL TECHNIQUES WITHIN CASE MEMORY 96
DISCUSSION 97
References 97
CHAPTER 28. A SCHEMA FOR AN INTEGRATED LEARNING SYSTEM 98
ABSTRACT 98
EMPIRICAL DATA VERSUS REASONING 98
ANALYSIS VERSUS SYNTHESIS 98
A PRIORI KNOWLEDGE VERSUS EMPIRICAL KNOWLEDGE 99
CONCLUSIONS 99
ACKNOWLEDGEMENTS 100
REFERENCES 100
CHAPTER 29. COMBINING EXPLANATION-BASED LEARNING AND ARTIFICIAL NEURAL NETWORKS 101
ABSTRACT 101
INTRODUCTION 101
AN INTEGRATED APPROACH 102
CONCLUSION 103
References 103

104 
CHAPTER 30. LEARNING CLASSIFICATION RULES USING BAYES 105
ABSTRACT 105
INTRODUCTION 105
THEORY 105
EXPERIMENTS 107
CONCLUSION 109
Acknowledgements 109
References 109
CHAPTER 31. NEW EMPIRICAL LEARNING MECHANISMS PERFORM SIGNIFICANTLY BETTER IN REAL LIFE DOMAINS 110
ABSTRACT 110
INTRODUCTION 110
REAL LIFE DOMAINS 110
EMPIRICAL LEARNING SYSTEMS 111
REDUNDANT KNOWLEDGE 111
GINESYS 112
EMPIRICAL TESTS 112
DISCUSSION 113
ACKNOWLEDGMENTS 113
REFERENCES 113
CHAPTER 32. INDUCTIVE LEARNING WITH BCT 115
ABSTRACT 115
INTRODUCTION 115
KNOWLEDGE REPRESENTATION 115
LEARNING OPERATORS 116
HEURISTIC MEASURES 116
ALGORITHM 117
EMPIRICAL RESULTS 117
DISCUSSION 118
CONCLUDING REMARKS 119
Acknowledgments 119
References 119
CHAPTER 33. WHAT GOOD ARE EXPERIMENTS? 120
ABSTRACT 120
INTRODUCTION 120
RESULTS 121
REFERENCES 123
Chapter 
124 
Abstract 124
1 Introduction 124
2 Definitions 125
3 Experiments 125
4 Discussion 128
References 129
CHAPTER 35. TWO ALGORITHMS THAT LEARN DNF BY 
130 
INTRODUCTION 130
DEFINITIONS 131
FRINGE ALGORITHM 132
GREEDY3 ALGORITHM 132
RESULTS 133
References 134
CHAPTER 36. LIMITATIONS ON INDUCTIVE LEARNING 135
ABSTRACT 135
INTRODUCTION 135
NOTATION 135
EXPERIMENTAL RESULTS 136
AN UPPER BOUND 137
IMPLICATIONS 139
BIBLIOGRAPHY 139
CHAPTER 37. THE INDUCTION OF PROBABILISTIC RULE SETS — THE ITRULE ALGORITHM 140
Abstract 140
Introduction 140
Motivation: why use sets of rules? 140
Previous work on learning sets of rules 141
The J-measure and the ITRULE algorithm 141
Inference using probabilistic rule sets 142
Acknowledgement 143
References 143
CHAPTER 38. EMPIRICAL SUBSTRUCTURE DISCOVERY 144
Abstract 144
1 Introduction 144
2 Substructure Discovery 144
3 Example 146
4 Conclusion 147
Acknowledgements 147
References 147
CHAPTER 39. LEARNING THE BEHAVIOR OF DYNAMICAL SYSTEMS FROM EXAMPLES 148
ABSTRACT 148
DYNAMICAL SYSTEMS 148
TOPOLOGICAL MAPS - A BRIEF DESCRIPTION 149
LEARNING MOVEMENTS OF A ROBOT ARM 149
CONCLUSION 151
ACKNOWLEDGEMENTS 151
REFERENCES 151
CHAPTER 40. EXPERIMENTS IN ROBOT LEARNING 152
INTRODUCTION 152
THE TASK DOMAIN 152
EXPERIMENTS WITH TWO LEARNING ROBOTS 153
FUTURE DIRECTIONS 154
References 155
Chapter 
157 
Abstract 157
1 Introduction 157
2 Shortcomings of ID3 158
3 The INFERULE Algorithm 159
4 Results 160
5 Conclusions 161
Acknowledgements 161
References 161
CHAPTER 42. KNOWLEDGE INTENSIVE INDUCTION 162
ABSTRACT 162
LEARNING DISJUNCTIVE CONCEPTS 162
BEYOND ID3 163
CONSTRAINING SEARCH USING FRAMES 164
EXTRACTING CONCEPT DESCRIPTIONS FROM THE DECISION TREE 165
CONCLUSION 166
Acknowledgments 166
References 166
CHAPTER 43. 
167 
ABSTRACT 167
INTRODUCTION 167
INDUCT: A STATISTICALLY WELL-FOUNDED EMPIRICAL INDUCTION PROCEDURE FOR DERIVING DECISION RULES FROM DATASETS 168
THE TRADE-OF BETWEEN KNOWLEDGE AND DATA 169
CONCLUSIONS 170
References 170
CHAPTER 44. SIGNAL DETECTION THEORY: VALUABLE TOOLS FOR EVALUATING INDUCTIVE LEARNING 171
ABSTRACT 171
SIGNAL DETECTION THEORY AND ROC CURVES 171
SIGNAL DETECTION THEORY 171
EVALUATING MONOTONE DNF RULES 172
COMPARISON OF CONNECTIONIST MODELS 172
DEFINING EVALUATION FUNCTIONS FOR GENETIC SEARCH 173
References 174
CHAPTER 45. UNKNOWN ATTRIBUTE VALUES IN INDUCTION 175
ABSTRACT 175
INTRODUCTION 175
DESCRIPTION OF DATASETS 175
DESCRIPTION OF APPROACHES 176
UNKNOWN VALUES WHEN PARTITIONING 177
UNKNOWN VALUES WHEN CLASSIFYING 177
UNKNOWN VALUES IN SELECTING TESTS 178
CONCLUSIONS 179
Acknowledgement 179
References 179
CHAPTER 46. 
180 
ABSTRACT 180
Introduction 180
ID3 and Back-Propagation 180
Training conventions 181
Behavioral Characterizations 181
Concluding Remarks 184
References 184
CHAPTER 47. BACON, DATA ANALYSIS AND ARTIFICIAL INTELLIGENCE 185
ABSTRACT 185
AN EXAMPLE 185
BACON VERSUS THE EVIDENCE 187
DOMAIN-INDEPENDENT DATA ANALYSIS 188
References 189
Part 3: Learning Plan Knowledge 190
CHAPTER 48. LEARNING TO PLAN IN COMPLEX DOMAINS 191
ABSTRACT 191
LEARNING SUBGOAL SEQUENCES FOR PLANNING 191
SUMMARY OF RESULTS 193
References 193
CHAPTER 49. AN EMPIRICAL ANALYSIS OF EBL APPROACHES FOR LEARNING PLAN SCHEMATA 194
ABSTRACT 194
SCHEMA-BASED PLANNING AND EXPLANATION-BASED LEARNING 194
EXPERIMENTAL METHODOLOGY 195
EXPERIMENTAL RESULTS AND DISCUSSION 196
CONCLUSION 198
Acknowledgements 198
References 198
CHAPTER 50. LEARNING DECISION RULES FOR SCHEDULING PROBLEMS: 
199 
ABSTRACT 199
INTRODUCTION 199
SCHEDULING PROBLEMS 199
THE PREDICATE CLASSIFIER SYSTEM 199
MINIMUM LATENESS SCHEDULING PROBLEMS 200
MINIMUM WEIGHTED TARDINESS SCHEDULING 200
CONCLUSION 201
REFERENCES 201
CHAPTER 51. LEARNING TACTICAL PLANS FOR PILOT AIDING 202
ABSTRACT 202
THE LEARNING DOMAIN AND PERFORMANCE PROBLEM 202
PERFORMANCE SYSTEM 202
REPRESENTATION OF INPUTS AND OUTPUTS OF LEARNING 202
LEARNING ALGORITHM 203
SUMMARY 204
REFERENCES 204
CHAPTER 52. ISSUES IN THE JUSTIFICATION-BASED DIAGNOSIS OF PLANNING FAILURES 205
Reference 207
CHAPTER 53. LEARNING PROCEDURAL KNOWLEDGE IN THE EBG CONTEXT 208
ABSTRACT 208
INTRODUCTION 208
THE LEARNING METHOD 209
DESCRIPTION OF THE LEARNING METHOD 209
RESULTS OF LEARNING 210
CONCLUSION 210
Acknowledgements 210
References 210
CHAPTER 54. LEARNING INVARIANTS FROM EXPLANATIONS 211
ABSTRACT 211
1. INTRODUCTION 211
2 OVERVIEW OF THE LIFE SYSTEM 212
3. REASONING ABOUT IMPOSSIBILITY 212
4 THE BLOCKING GRAPH 213
5 THE FRAME PROBLEM 214
6 CONCLUSIONS 214
Acknowledgements 215
References 215
Chapter 
216 
Abstract 216
1 INTRODUCTION 216
2 METHODOLOGY 216
3 EXAMPLE PROBLEM 217
4 CONCLUSIONS 219
References 219
CHAPTER 56. LEARNING TO RECOGNIZE PLANS INVOLVING 
220 
INTRODUCTION 220
THE SYSTEM 220
AN EXAMPLE 220
CONCLUSION CURRENT AND FUTURE WORK
References 222
Chapter 
223 
Eureka's components 223
Accounting for psychological phenomena 224
Current status of the model 225
References 225
CHAPTER 58. Discovering problem solving strategies: 
226 
The experiment and the protocol 226
Classification of the learning events 226
Conclusions 228
References 228
Chapter 
229 
1 Introduction 229
2 Our Approach 229
3 Experiments 230
4 Discussion 231
References 231
Chapter 
232 
1 Introduction 232
2 Lazy Explanation-Based Learning 232
3 Knowledge Enabled Planning 233
4 Conclusion 234
Acknowledgments 234
References 234
CHAPTER 61. LEARNING APPROXIMATE PLANS FOR USE IN THE REAL WORLD 235
ABSTRACT 235
INTRODUCTION 235
THE MODEL 236
AN EXAMPLE 238
RELATED WORK AND CONCLUSIONS 239
REFERENCES 239
Chapter 
240 
1. Introduction 240
2. Representation and Planning in Daedalus 240
3. Acquiring and Using Plan Knowledge 241
5. Behavior of Daedalus 241
References 242
CHAPTER 63. Conceptual Clustering of Mean-Ends Plans 243
ABSTRACT 243
INTRODUCTION 243
CONCEPTUAL CLUSTERING OF OPERATORS 243
PLAN GENERATION and REUSE 244
CONCLUDING REMARKS 245
Acknowledgements 245
References 245
CHAPTER 64. LEARNING APPROPRIATE ABSTRACTIONS FOR PLANNING IN FORMATION PROBLEMS 246
ABSTRACT 246
Introduction 246
The PLACE system 247
Learning new abstractions 249
Concluding remarks 250
References 250
Chapter 
251 
Abstract 251
CHAPTER 66. LEARNING HIERARCHIES OF ABSTRACTION SPACES 252
ABSTRACT 252
INTRODUCTION 252
ABSTRIPS 252
ALPINE 253
PROPERTIES OF ABSTRACT PLANS 254
THE LEARNING METHOD 255
CONCLUSIONS 256
Acknowledgments 256
References 256
CHAPTER 67. LEARNING FROM OPPORTUNITY 257
ABSTRACT 257
PLANNING AND LEARNING 257
OPPORTUNISTIC MEMORY 257
AN EXAMPLE 258
CONCLUSION 259
REFERENCES 259
CHAPTER 68. LEARNING BY ANALYZING FORTUITOUS OCCURRENCES 260
ABSTRACT 260
INTRODUCTION 260
DETECTION 260
REFINEMENT 261
AN EXAMPLE 261
DISCUSSION AND CONCLUSION 262
ACKNOWLEDGEMENTS 262
REFERENCES 262
CHAPTER 69. EXPLANATION-BASED LEARNING OF REACTIVE OPERATORS 263
ABSTRACT 263
INTRODUCTION 263
REACTIVITY IN PLANNING 263
REACTIVE OPERATORS 264
DISCUSSION 265
ACKNOWLEDGMENTS 265
REFERENCES 265
CHAPTER 70. ON BECOMING REACTIVE 266
INTRODUCTION 266
LEARNING 267
DISCUSSION 267
Acknowledgements 268
References 268
Part 4: 
270 
CHAPTER 71. KNOWLEDGE BASE REFINEMENT AND THEORY REVISION 271
INTRODUCTION 271
THEORIES AND EXPERT SYSTEMS 271
THEORETICAL TERMS AND BRIDGE LAWS 272
GOALS OF THEORY REVISION 273
ELEMENTS OF THEORY REVISION 274
REDUCTION OF THEORETICAL TERMS AND THEORY REVISION 275
Acknowledgments 276
References 276
CHAPTER 72. THEORY FORMATION BY ABDUCTION: 
277 
ABSTRACT 277
ABDUCTION, HYPOTHESIS FORMATION, AND THEORY REVISION 277
THE CHEMICAL REVOLUTION 278
SOME ASPECTS OF THE PHLOGISTON THEORY ENCODED AS RULES 278
ABDUCTION OF ASPECTS OF THE OXYGEN THEORY 280
CONCLUSION 281
Acknowledgments 282
References 282
CHAPTER 73. USING DOMAIN KNOWLEDGE TO AID SCIENTIFIC THEORY REVISION 283
ABSTRACT 283
INTRODUCTION 283
AN OVERVIEW OF THE REVOLVER SYSTEM 283
ADDING DOMAIN KNOWLEDGE TO THE EVALUATION FUNCTION 284
EVALUATING REVOLVER'S LEARNING BEHAVIOR 287
DISCUSSION 288
References 288
Chapter 
289 
Abstract 289
1. Motivation 289
2. Question addressed in this research 289
3. Methodology 290
4. The Structure of KEKADA 290
5. Evaluation of KEKADA performance 292
6. Conclusion 293
7. Acknowledgement 293
References 293
CHAPTER 75. EXEMPLAR-BASED THEORY REJECTION: AN APPROACH TO THE EXPERIENCE CONSISTENCY 
295 
Abstract 295
1 Introduction 295
2 Explanation-based Theory Revision - An Overview 296
3 Exemplar-based Theory Rejection 296
4 Discussion 300
References 300
CHAPTER 76. CONTROLLING SEARCH FOR THE CONSEQUENCES OF NEW INFORMATION DURING KNOWLEDGE INTEGRATION 301
ABSTRACT 301
INTRODUCTION 301
KI: A TOOL FOR KNOWLEDGE INTEGRATION 302
CONTROLLING THE SEARCH FOR CONSEQUENCES 305
SUMMARY 306
References 306
CHAPTER 77. IDENTIFYING KNOWLEDGE BASE DEFICIENCIES BY OBSERVING USER BEHAVIOR 307
ABSTRACT 307
INTRODUCTION 307
THE PERFORMANCE PROBLEM: AN INTELLIGENT AUTOMATED PILOT'S ASSISTANT 307
REFINEMENT OF THE PA KNOWLEDGE BASE 308
REFINEMENT OF THE EBL DOMAIN THEORY 309
SUMMARY 311
REFERENCES 311
Chapter 
313 
1 Introduction 313
2 A Boolean Function Design System 313
3 A System Development Sequence 315
4 Conclusions 318
Acknowledgements 318
CHAPTER 79. DISCOVERING MATHEMATICAL OPERATOR DEFINITIONS 319
ABSTRACT 319
INTRODUCTION 319
GENERATE, PRUNE, AND PROVE METHOD 321
CONCLUDING REMARKS 323
References 324
CHAPTER 80. IMPRECISE CONCEPT LEARNING WITHIN A GROWING LANGUAGE 325
ABSTRACT 325
INTRODUCTION 325
BASIC DEFINITIONS 326
LEARNING PROCESS 328
References 330
CHAPTER 81. USING DETERMINATIONS IN EBL: A SOLUTION TO THE INCOMPLETE THEORY PROBLEM 331
ABSTRACT 331
1 Introduction 331
2 Determinations 332
3 One View of the Incomplete Theory Problem 332
4 A Technique for Refining Incomplete Theories 333
5 Conclusions 335
Acknowledgements 336
References 336
Chapter 
337 
1. Introduction 337
2. Learning protocols and a model of a rule base 337
3. Some NP-Complete refinement problems 338
4. Rule-based systems as classifiers 340
5. Gradualness 341
6. Conclusions 342
Acknowledgements 342
References 342
CHAPTER 83. KNOWLEDGE BASE REFINEMENT AS IMPROVING AN 
343 
ABSTRACT 343
INTRODUCTION 343
ProHC HEURISTIC CLASSIFICATION SHELL 343
INCORRECT DOMAIN THEORY 345
INCONSISTENT DOMAIN THEORY 345
INCOMPLETE DOMAIN THEORY 345
EXPERIMENTAL RESULTS 347
SUMMARY 
348 
Acknowledgements 348
References 348
Part 5: Incremental Learning 350
CHAPTER 84. INCREMENTAL LEARNING OF CONTROL STRATEGIES WITH GENETIC ALGORITHMS 351
INTRODUCTION 351
THE EVASIVE MANEUVERS PROBLEM 351
THE PERFORMANCE MODULE: CPS 352
THE LEARNING MODULE 353
A CASE STUDY 354
DISCUSSION 354
References 355
CHAPTER 85. TOWER OF HANOI WITH CONNECTIONIST NETWORKS: 
356 
ABSTRACT 356
INTRODUCTION 356
CREDIT ASSIGNMENT 356
THE NETWORKS 357
RESULTS 358
DISCUSSION 359
CONCLUSION 359
References 360
Chapter 
361 
Learning to Act 361
Performance Criteria 362
Related Work 364
Acknowledgments 364
References 364
Chapter 
365 
1 Acting and Reasoning 365
2 The Reactive Component 366
3 Adding Projection 367
4 Discussion 368
References 368
CHAPTER 88. UNCERTAINTY BASED SELECTION OF LEARNING EXPERIENCES 369
ABSTRACT 369
STRATEGIES FOR SELECTING EXPERIENCES 369
UNCERTAINTY AS AN EXPERIENCE SELECTION HEURISTIC 370
UNCERTAINTY BASED EXPERIENCE SELECTION IN DIDO 371
DISCUSSION 372
References 372
CHAPTER 89. IMPROVED TRAINING VIA INCREMENTAL LEARNING 373
ABSTRACT 373
INTRODUCTION 373
INSTANCE SELECTION STRATEGIES OF ID3 AND ID5R 373
AN EXPERIMENT 374
CONCLUSION 375
Acknowledgements 376
References 376
CHAPTER 90. INCREMENTAL BATCH LEARNING 377
INTRODUCTION 377
INCREMENTAL BATCH LEARNING TESTBED 377
INFLUENCE OF BATCH SIZE 378
INSTANCE STORAGE AND CONCEPT DRIFT 379
COMPARISON OF INCREMENTAL BATCH LEARNING METHODS 379
TIME CONSTRAINTS, NOISY AND ANOMALOUS DATA 381
CONCLUSIONS 381
References 381
CHAPTER 91. INCREMENTAL CONCEPT FORMATION WITH COMPOSITE OBJECTS 382
1. INTRODUCTION 382
2. REPRESENTATION AND ORGANIZATION IN LABYRINTH 382
3. CLASSIFICATION AND LEARNING IN LABYRINTH 383
4. DISCUSSION 384
Acknowledgment 385
References 385
CHAPTER 92. 
386 
ABSTRACT 386
INTRODUCTION 386
BACKGROUND 386
MULTIPLE REPRESENTATIONS 387
DISCUSSION 389
Acknowledgements 389
References 389
CHAPTER 93. FOCUSED CONCEPT FORMATION 390
1. Introduction 390
2. Previous Work 390
3. An Incremental Algorithm for Attention 391
4. Performance Tasks and Planned Experiments 392
5. Summary 393
Acknowlegments 393
References 393
CHAPTER 94. An Exploration into Incremental Learning: the INFLUENCE system 394
Abstract 394
1. INTRODUCTION 394
2. THE INFLUENCE SYSTEM 395
3. EXPERIMENTS 396
4. CONCLUSION 397
Acknowledgements 397
References 397
CHAPTER 95. INCREMENTAL, INSTANCE-BASED LEARNING OF INDEPENDENT AND GRADED CONCEPT DESCRIPTIONS 398
ABSTRACT 398
1. MOTIVATION 398
2. THE INSTANCE-BASED PROCESS FRAMEWORK 398
3. BLOOM: LEARNING CONCEPT-DEPENDENT ATTRIBUTE WEIGHTS 399
4. EMPIRICAL STUDIES AND RESULTS 400
5. LIMITATIONS AND SUMMARY 402
Acknowledgements 402
References 402
Chapter 
403 
1 Introduction 403
2 Building and using cost decision trees 403
3 Learning cost versus application cost 405
4 Tradeoff among environmental costs 406
Acknowledgements 406
References 406
CHAPTER 97. REDUCING REDUNDANT LEARNING 407
ABSTRACT 407
INTRODUCTION 407
NONREDUNDANT LEARNING 407
CORA 408
CORA'S EMPIRICAL BEHAVIOR 408
SUMMARY AND FUTURE DIRECTIONS 410
References 410
CHAPTER 98. INCREMENTAL CLUSTERING BY MINIMIZING REPRESENTATION LENGTH 411
ABSTRACT 411
Introduction 411
Cluster configuration quality 411
Incremental clustering strategy 412
Implementation and test results 412
References 414
CHAPTER 99. INFORMATION FILTERS AND THEIR 
415 
INTRODUCTION 415
INFORMATION FILTERING 415
INFORMATION FILTERS IN THE SYLLOG SYSTEM 416
CONCLUSIONS 418
References 418
CHAPTER 100. ADAPTIVE LEARNING OF DECISION-THEORETIC 
419 
ABSTRACT 419
DTA: DECISION-THEORETIC REAL-TIME HEURISTIC SEARCH 419
INCREMENTAL LEARNING OF PARAMETERS 421
CONCLUSIONS 422
Acknowledgments 422
References 422
CHAPTER 101. 
423 
ABSTRACT 423
INTRODUCTION: ADAPTIVE STRATEGIES IN ZERO-SUM GAMES 423
ADAPTATION AND LEARNING 423
EXPERIMENTS AND DISCUSSION 425
BIBLIOGRAPHY 426
CHAPTER 102. AN INCREMENTAL GENETIC ALGORITHM FOR REAL-TIME LEARNING 427
ABSTRACT 427
INTRODUCTION 427
THE GENETIC ALGORITHMS 427
THE EXPERIMENTS 428
THE RESULTS 429
References 430
CHAPTER 103. PARTICIPATORY LEARNING: A CONSTRUCTIVIST MODEL 431
INTRODUCTION 431
A CONSTRUCTIVIST MODEL OF PARTICIPATORY LEARNING 431
REFERENCES 434
Part 6: 
436 
CHAPTER 104. REPRESENTATIONAL ISSUES IN MACHINE LEARNING 437
OBJECTIVES 437
ISSUES AND APPROACHES 437
FUTURE DIRECTIONS 439
References 440
CHAPTER 105. Labor Saving New Distinctions 441
Abstract 441
1 Introduction 441
2 Framework 441
3 New Relations 441
4 Complexity of Introducing New Relations 443
5 New Objects 443
6 Conclusions 444
Acknowledgements 444
References 444
CHAPTER 106. A THEORY OF JUSTIFIED REFORMULATIONS 445
ABSTRACT 445
REFORMULATION 445
JUSTIFYING REFORMULATIONS 446
IRRELEVANCE REFORMULATIONS 447
THE RELEVANCE OF IRRELEVANCE 448
CONCLUSIONS 449
References 449
CHAPTER 107. REFORMULATION PROM STATE SPACE TO REDUCTION SPACE 450
INTRODUCTION 450
THE PROBLEM 450
THE SOLUTION 451
REFERENCES 451
CHAPTER 108. KNOWLEDGE-BASED FEATURE GENERATION 452
ABSTRACT 452
INTRODUCTION 452
CREATING NEW TERMS FROM SEARCH PROBLEM SPECIFICATIONS 452
FUTURE WORK 454
CONCLUSIONS 454
Acknowledgement 454
References 454
CHAPTER 109. ENRICHING VOCABULARIES BY GENERALIZING EXPLANATION STRUCTURES 455
ABSTRACT 455
INTRODUCTION 455
PLEESE: AN AUTOMATIC PROGRAMMING SYSTEM 456
CURRENT RESEARCH DIRECTIONS 457
CONCLUSION 457
References 457
Chapter 
458 
Abstract 458
1 Introduction 458
2 Higher-order EBG and Representation Domains 459
3 EBG and Modal Logic 459
References 460
CHAPTER 111. 
461 
1 Motivation 461
2 Framework 462
3 Finding the Optimal Strategy in a Redundant KB 463
4 Conclusion 464
References 464
CHAPTER 112. A MATHEMATICAL FRAMEWORK FOR STUDYING REPRESENTATION 465
ABSTRACT 465
MOTIVATION 465
INTUITIONS ABOUT THE NATURE OF REPRESENTATION 465
FORMALIZING "REPRESENTATION" 466
References 467
Chapter 
468 
1 Introduction 468
2 Changing Representations with Bumble 469
3 The Good News and The Bad News 470
Acknowledgements 471
References 471
CHAPTER 114. COMPARING SYSTEMS AND ANALYZING FUNCTIONS TO IMPROVE CONSTRUCTIVE INDUCTION 472
1. INTRODUCTION: CONSTRUCTION IN CONCEPT LEARNING 472
2. ONE BENEFIT OF JUXTAPOSING TWO SYSTEMS 473
3. BROAD BENEFITS OF COMPARING MANY SYSTEMS 474
References 475
CHAPTER 115. EVALUATING ALTERNATIVE INSTANCE REPRESENTATIONS 476
ABSTRACT 476
INTRODUCTION 476
WHAT IS A GOOD REPRESENTATION? 477
CHOOSING A GOOD REPRESENTATION 477
EXPERIMENT 478
CONCLUSIONS 478
Acknowledgement 479
References 479
CHAPTER 116. EVALUATING BIAS DURING PAC-LEARNING 480
INTRODUCTION 480
PAC-LEARNING 480
DETECTING INCORRECT BIAS 480
RELATED WORK 481
CONCLUSION 482
Acknowledgements 482
References 482
CHAPTER 117. CONSTRUCTING REPRESENTATIONS USING INVERTED SPACES 483
ABSTRACT 483
INTRODUCTION 483
LINEAR SEPARABILITY 483
FEATURE CONSTRUCTION 484
INTEGRATION WITH DIMENSIONALITY REDUCTION TECHNIQUES 484
FUTURE WORK 484
References 484
CHAPTER 118. A CONSTRUCTIVE INDUCTION FRAMEWORK 485
INTRODUCTION 485
THE FRAMEWORK 485
CHAPTER 119. CONSTRUCTIVE INDUCTION BY ANALOGY 487
ABSTRACT 487
THE TECHNIQUE 487
References 488
CHAPTER 120. CONCEPT DISCOVERY THROUGH UTILIZATION OF INVARIANCE EMBEDDED IN THE DESCRIPTION LANGUAGE 489
ABSTRACT 489
THE LEARNING PROBLEM 489
INVARIANCE AND CONSTRUCTIVE INDUCTION 489
THE CONSTRUCTIVE INDUCTION PROCEDURE 490
CONCLUSIONS 490
REFERENCES 490
CHAPTER 121. DECLARATIVE BIAS FOR STRUCTURAL DOMAINS 491
ABSTRACT 491
Situation Identification and Declarative Bias 491
Structural Domains 491
Isomorphic Determinations 492
Discussion 493
Acknowledgments 493
References 493
Chapter 
494 
References 495
Chapter 
496 
1 Introduction 496
2 Programming by Informing 496
3 A Knowledge-level Description of Agents 497
4 Analysis 497
5 Conclusion 499
References 499
CHAPTER 124. AN OBJECT-ORIENTED REPRESENTATION FOR SEARCH ALGORITHMS 500
ABSTRACT 500
INTRODUCTION 500
AN OBJECT-ORIENTED REPRESENTATION FOR GENERATORS 500
CURRENT STATUS 502
References 502
CHAPTER 125. COMPILING LEARNING VOCABULARY FROM A PERFORMANCE SYSTEM DESCRIPTION 503
ABSTRACT 503
INTRODUCTION 503
A LANGUAGE FOR DESCRIBING SOLVER'S BEHAVIOR 503
SYNTHESIZING THE "USEFUL" PREDICATE 504
CONCLUSIONS 506
References 506
CHAPTER 126. GENERALIZED RECURSIVE SPLITTING ALGORITHMS FOR LEARNING HYBRID CONCEPTS 507
ABSTRACT 507
INTRODUCTION 507
GENERALIZED RECURSIVE SPLITTING ALGORITHMS 507
THE CRL ALGORITHM 508
EXPERIMENTAL RESULTS 508
CONCLUSIONS 508
Acknowledgements 509
References 509
CHAPTER 127. SCREENING HYPOTHESES WITH EXPLICIT BIAS 510
INTRODUCTION 510
PREDICTOR: A SYSTEM THAT USES EXPLICIT BIAS 510
Acknowledgements 511
References 511
CHAPTER 128. BUILDING A LEARNING BIAS FROM PERCEIVED DEPENDENCIES 512
INTRODUCTION 512
THE NOTION OF A PERCEIVED DEPENDENCY 512
THE LEARNING BIAS 512
BUILDING A LEARNING BIAS USING PERCEIVED DEPENDENCIES 513
CONCLUSION 513
Reference 513
CHAPTER 129. A BOOTSTRAPPING APPROACH TO CONCEPTUAL CLUSTERING 514
Text 514
CHAPTER 130. OVERCOMING FEATURE SPACE BIAS IN A REACTIVE ENVIRONMENT 516
ABSTRACT 516
THE SEMANTICS OF INDEFINABLE TERMS IN SELF-REFLECTIVE ARCHITECTURES 516
OVERVIEW OF NOME'S ENVIRONMENT 516
DIMENSIONS AND THE DIMENSION STUDIER 516
EVALUATING DIMENSIONAL I/O PAIRINGS AGAINST THE EXTERNAL WORLD 518
ACKNOWLEDGMENTS 518
References 518
AUTHOR INDEX 520

Produkt-Details

• Erscheint lt. Verlag: 28.6.2014
• Sprache: englisch
• Themenwelt: Informatik ► Theorie / Studium ► Künstliche Intelligenz / Robotik
• ISBN-10: 1-4832-9740-3 / 1483297403
• ISBN-13: 978-1-4832-9740-8 / 9781483297408