Machine Learning Proceedings 1995 (eBook)

Proceedings of the Twelfth International Conference on Machine Learning, Tahoe City, California, July 9-12 1995

Armand Prieditis, Stuart Russell (Herausgeber)

eBook Download: PDF

2014 | 1. Auflage
400 Seiten
Elsevier Science (Verlag)
978-1-4832-9866-5 (ISBN)

Infos

Machine Learning Proceedings 1995

Inhalt

Front Cover 1
Machine Learning 2
Copyright Page 3
Table of Contents 4
Preface 10
Advisory Committee 11
Program Committee 11
Auxiliary Reviewers 12
Workshops 12
Tutorials 12
PART 1: CONTRIBUTED PAPERS 16
Chapter 1. On-line Learning of Binary Lexical Relations Using Two-dimensional Weighted Majority Algorithms 18
ABSTRACT 18
1 Introduction 18
2 On-line Learning Model for Binary Relations 20
3 Two-dimensional Weighted Majority Prediction Algorithms 20
4 Experimental Results 21
5 Theoretical Performance Analysis 23
6 Concluding Remarks 26
Acknowledgement 26
References 26
Chapter 2. On Handling Tree-Structured Attributes in Decision Tree Learning 27
Abstract 27
1 Introduction 27
2 Decision Trees With Tree-Structured Attributes 28
3 Pre-processing Approaches 29
4 A Direct Approach 30
5 Analytical Comparison 31
6 Experimental Comparison 33
7 Summary and Conclusion 34
Acknowledgement 35
References 35
Chapter 3. Theory and Applications of Agnostic PAC-Learning with Small Decision Trees 36
Abstract 36
1 INTRODUCTION 36
2 THE AGNOSTIC PAC-LEARNING ALGORITHM T2 38
3 EVALUATION OF T2 ON "REAL-WORLD" CLASSIFICATION PROBLEMS 40
4 LEARNING CURVES FOR DECISION TREES OF SMALL DEPTH 42
5 CONCLUSION 43
Acknowledgement 43
References 44
Chapter 4. Residual Algorithms: Reinforcement Learning with Function Approximation 45
ABSTRACT 45
1 INTRODUCTION 45
2 ALGORITHMS FOR LOOKUP TABLES 46
3 DIRECT ALGORITHMS 46
4 RESIDUAL GRADIENT ALGORITHMS 47
5 RESIDUAL ALGORITHMS 48
6 STOCHASTIC MDPS AND MODELS 50
7 MDPS WITH MULTIPLE ACTIONS 50
8 RESIDUAL ALGORITHM SUMMARY 50
9 SIMULATION RESULTS 51
10 CONCLUSIONS 52
Acknowledgments 52
References 52
Chapter 5. Removing the Genetics from the Standard Genetic Algorithm 53
Abstract 53
1. THE GENETIC ALGORITHM (GA) 53
2. FOUR PEAKS: A PROBLEM DESIGNED TO BE GA-FRIENDLY 54
3. SELECTING THE GA'S PARAMETERS 55
4. POPULATION-BASED INCREMENTAL LEARNING 56
5. EMPIRICAL ANALYSIS ON THE FOUR PEAKS PROBLEM 57
6. DISCUSSION 59
7. CONCLUSIONS 60
ACKNOWLEDGEMENTS 60
REFERENCES 60
Chapter 6. Inductive Learning of Reactive Action Models 62
Abstract 62
1 INTRODUCTION 62
2 CONTEXT OF THE LEARNER 62
3 ACTIONS AND TELEO-OPERATORS 63
4 COLLECTING INSTANCES FOR LEARNING 64
5 THE INDUCTIVE LOGIC PROGRAMMING ALGORITHM 65
6 EVALUATION 66
7 RELATED WORK 67
8 FUTURE WORK 68
Acknowledgements 68
References 68
Chapter 7. Visualizing High-Dimensional Structure with the Incremental Grid Growing Neural Network 70
Abstract 70
1 INTRODUCTION 70
2 INCREMENTAL GRID GROWING 71
3 COMPARISON USING MINIMUM SPANNING TREEDATA 73
4 DEMONSTRATION USING REALWORLD SEMANTIC DATA 73
5 DISCUSSION AND FUTURE WORK 75
6 CONCLUSION 77
References 77
Chapter 8. Empirical support for Winnow and Weighted-Majority based algorithms: results on a calendar scheduling domain 79
Abstract 79
1 Introduction 79
2 The learning problem 80
3 Description of the algorithms 80
4 Experimental results 82
5 Theoretical results 85
Acknowledgements 87
References 87
Appendix 87
Chapter 9. Automatic Selection of Split Criterion during Tree Growing Based on Node Location 88
Abstract 88
1 DECISION TREE CONSTRUCTION 88
2 SITUATIONS IN WHICH ACCURACY IS THE BEST SPLITCRITERION 89
3 IMPLICATIONS FOR TREE-GROWING ALGORITHMS 90
4 EMPIRICAL SUPPORT OF THE HYPOTHESIS 90
5 FUTURE DIRECTIONS 94
References 94
Chapter 10. A Lexically Based Semantic Bias for Theory Revision 96
Abstract 96
1 INTRODUCTION 96
2 BACKGROUND 97
3 CLARUS 97
4 RESULTS 100
5 Discussion 103
6 CONCLUSION 104
Acknowledgments 104
References 104
Chapter 11. A Comparative Evaluation of Voting and Meta-learning on Partitioned Data 105
Abstract 105
1 Introduction 105
2 Common Voting and Statistical Techniques 105
3 Meta-learning Techniques 106
4 Experiments and Results 107
5 Arbiter Tree 110
6 Discussion 112
7 Concluding Remarks 112
References 113
Chapter 12. Fast and Efficient Reinforcement Learning with Truncated Temporal Differences 114
Abstract 114
1 INTRODUCTION 114
2 TD-BASED ALGORITHMS 115
3 TRUNCATED TEMPORAL DIFFERENCES 116
4 EXPERIMENTAL STUDIES 120
5 CONCLUSION 120
Acknowledgements 122
References 122
Chapter 13. K*: An Instance-based Learner Using an Entropie Distance Measure 123
Abstract 123
1 INTRODUCTION 123
2 ENTROPY AS A DISTANCE MEASURE 124
3 K* ALGORITHM 127
4 RESULTS 128
5 CONCLUSIONS 129
Acknowledgments 129
References 129
Chapter 14. Fast Effective Rule Induction 130
Abstract 130
1 INTRODUCTION 130
2 PREVIOUS WORK 130
3 EXPERIMENTS WITH IREP 132
4 IMPROVEMENTS TO IREP 134
5 CONCLUSIONS 137
References 138
Chapter 15. Chapter Text Categorization and Relational Learning 139
Abstract 139
1 INTRODUCTION 139
2 TEXT CATEGORIZATION 139
3 AN EXPERIMENTAL TESTBED 140
4 THE LEARNING METHOD 140
5 EVALUATING THERELATIONAL ENCODING 141
6 RELATION SELECTION 143
7 MONOTONICITY CONSTRAINTS 144
8 COMPARISON TO OTHER METHODS 145
9 CONCLUSIONS 146
Acknowledgements 146
References 147
Chapter 16. Protein Folding: Symbolic Refinement Competes with Neural Networks 148
Abstract 148
1 INTRODUCTION 148
2 THE PROTEIN FOLDING DOMAIN 148
3 RELATED WORK 150
4 KRUST'S SYMBOLIC REFINEMENT 151
5 EXPERIMENTAL RESULTS 153
6 SUMMARY 155
References 156
Chapter 17. A Bayesian Analysis of Algorithms for Learning Finite Functions 157
Abstract 157
1 Introduction 157
2 Preliminaries 158
3 Algorithms and priors 159
4 Approaches to prior and algorithm selection 161
5 Discussion and future work 162
Acknowledgements 164
References 164
Chapter 18. Committee-Based Sampling For Training Probabilistic Classifiers 165
Abstract 165
1 INTRODUCTION 165
2 BACKGROUND 166
3 COMMITTEE-BASEDSAMPLING 167
4 HMMS AND PART-OF-SPEECHTAGGING 168
5 COMMITTEE-BASEDSAMPLING FOR HMMS 168
6 EXPERIMENTAL RESULTS 170
7 CONCLUSIONS 171
References 171
Chapter 19. Learning Prototypical Concept Descriptions 173
Abstract 173
1 INTRODUCTION 173
2 LEARNING PROTOTYPICALDESCRIPTIONS 174
3 EVALUATION 176
4 DISCUSSION AND FUTUREDIRECTIONS 180
Acknowledgments 181
References 181
Chapter 20. A Case Study of Explanation-Based Control 182
Abstract 182
1 INTRODUCTION 182
2 THE ACROBOT 182
3 THE EBC APPROACH 183
4 A CONTROL THEORY SOLUTION 186
5 THE EBC SOLUTION 186
6 EMPIRICAL EVALUATION 188
7 CONCLUSIONS 189
Acknowledgements 190
References 190
Chapter 21. Explanation-Based Learning and Reinforcement Learning: A Unified View 191
Abstract 191
1 Introduction 191
2 Methods 193
3 Experiments and Results 196
4 Discussion 198
5 Conclusion 199
Acknowledgements 199
References 199
Chapter 22. Lessons from Theory Revision Applied to Constructive Induction 200
Abstract 200
1 Introduction 200
2 Context and Related Work 201
3 Demonstrations of Related Work 202
4 Theory-Guided Constructive Induction 205
5 Experiments 206
6 Discussion 207
References 208
Chapter 23. Supervised and Unsupervised Discretization of Continuous Features 209
Abstract 209
1 Introduction 209
2 Related Work 210
3 Methods 212
4 Results 213
5 Discussion 213
6 Summary 216
References 216
Chapter 24. Bounds on the Classification Error of the Nearest Neighbor Rule 218
Abstract 218
1 INTRODUCTION 218
2 DEFINITIONS AND THEOREMS 219
3 DISCUSSION AND CONCLUSION 222
Acknowledgements 222
References 222
Chapter 25. Q-Learning for Bandit Problems 224
Abstract 224
1 INTRODUCTION 224
2 BANDIT PROBLEMS 225
3 THE GITTINS INDEX 226
4 RESTART-IN-STATE-i PROBLEMS AND THE GITTINSINDEX 227
5 ON-LINE ESTIMATION OFGITTINS INDICES VIAQ-LEARNING 228
6 EXAMPLES 229
7 CONCLUSION 231
Acknowledgements 232
References 232
Chapter 26. Distilling Reliable Information From Unreliable Theories 233
Abstract 233
1 INTRODUCTION 233
2 IDENTIFYING STABLE EXAMPLES 233
3 USING STABILITY TO ELIMINATE NOISE 236
4 RESULTS 237
5 DISCUSSION 238
Acknowledgements 239
References 239
Chapter 27. A Quantitative Study of Hypothesis Selection 241
Abstract 241
1 Introduction 241
2 The Hypothesis Selection Problem 242
3 PAO Algorithms for Hypothesis Selection 242
4 Trading Off Exploitation and Exploration 245
5 Implication to Probabilistic Hill-Climbing 247
6 Related Work 248
7 Conclusion 248
Acknowledgements 249
References 249
Chapter 28. Learning proof heuristics by adapting parameters 250
Abstract 250
1 INTRODUCTION 250
2 FUNDAMENTALS 251
3 LEARNING PARAMETERS WITH A GA 252
4 THE UKB-PROCEDURE 253
5 DESIGNING A FITNESS FUNCTION 254
6 EXPERIMENTAL RESULTS 256
7 DISCUSSION 257
Acknowledgements 258
References 258
Chapter 29. Efficient Algorithms for Finding Multi-way Splits for Decision Trees 259
Abstract 259
1 Introduction 259
2 Computing Multi-Split Partitions 260
3 Experiments 262
4 Conclusion 265
Acknowledgements 266
References 266
Chapter 30. Ant-Q: A Reinforcement Learning approach to the traveling salesman problem 267
Abstract 267
1 INTRODUCTION 267
2 THE ANT-Q FAMILY OF ALGORITHMS 267
3 AN EXPERIMENTAL COMPARISONOF ANT-Q ALGORITHMS 268
4. TWO INTERESTING PROPERTIES OF ANT-Q 271
5 COMPARISONS WITH OTHER HEURISTICS AND SOME RESULTS ON DIFFICULT PROBLEMS 273
6 CONCLUSIONS 273
Acknowledgements 275
References 275
Chapter 31. Stable Function Approximation in Dynamic Programming 276
Abstract 276
1 INTRODUCTION AND BACKGROUND 276
2 DEFINITIONS AND BASIC THEOREMS 277
3 MAIN RESULTS: DISCOUNTED PROCESSES 278
4 NONDISCOUNTED PROCESSES 279
5 CONVERGING TO WHAT 281
6 EXPERIMENTS: HILL-CAR THE HARD WAY 281
7 CONCLUSIONS AND FURTHER RESEARCH 282
References 282
Chapter 32. The Challenge of Revising an Impure Theory 284
Abstract 284
1 Introduction 284
2 Framework 285
3 Computational Complexity 287
4 Prioritizing Default Theories 289
5 Conclusion 290
References 291
Chapter 33. Symbiosis in Multimodal Concept Learning 293
Abstract 293
1 INTRODUCTION 293
2 NICHE TECHNIQUES 294
3 SYSTEM OVERVIEW 294
4 INDIVIDUAL AND GROUP OPERATORS 296
5 FITNESS FUNCTION 297
6 COMPARISONS TO OTHER SYSTEMS 297
7 RESULTS 298
8 CONCLUSIONS 299
Acknowledgements 299
References 300
Chapter 34. Tracking the Best Expert 301
Abstract 301
1 INTRODUCTION 301
2 PRELIMINARIES 303
3 THE ALGORITHMS 303
4 FIXED SHARE ANALYSIS 304
5 VARIABLE SHARE ANALYSIS 305
6 EXPERIMENTAL RESULTS 308
References 309
Chapter 35. Reinforcement Learning by Stochastic Hill Climbing on Discounted Reward 310
Abstract 310
1 Introduction 310
2 Domain 310
3 Difficulties of Q-learning 312
4 Hill Climbing for Reinforcement Learning 312
5 Experiments 314
6 Discussion 316
7 Conclusion 316
Appendix 317
References 318
Chapter 36. Automatic Parameter Selection by Minimizing Estimated Error 319
Abstract 319
1 Introduction 319
2 The Parameter Selection Problem 320
3 The Wrapper Method 321
4 Automatic Parameter Selection for C4.5 322
5 Experiments with C4.5-AP 322
6 Related Work 325
7 Conclusion 326
Acknowledgments 326
References 326
Chapter 37. Error-Correcting Output Coding Corrects Bias and Variance 328
Abstract 328
1 Introduction 328
2 Definitions and Previous Work 329
3 Decomposing the Error Rate into Bias and Variance Components 331
4 ECOC and Voting 332
5 ECOC Reduces Variance and Bias 334
6 Bias Differences are Caused by Non-Local Behavior 334
7 Discussion and Conclusions 335
Acknowledgements 336
References 336
Chapter 38. Learning to Make Rent-to-Buy Decisions with Systems Applications 337
Abstract 337
1 Introduction 337
2 Definitions and Main Analytical Results 339
3 Algorithm Ae 339
4 Analysis 340
5 Adaptive Disk Spindown andRent-to-Buy 343
6 Experimental Results 343
Acknowledgements 344
References 344
Chapter 39. NewsWeeder: Learning to Filter Netnews 346
Abstract 346
1 INTRODUCTION 346
2 APPROACH 347
3 RESULTS 350
4 CONCLUSION 353
5 FUTURE WORK 353
Acknowledgments 353
References 353
Chapter 40. Hill Climbing Beats Genetic Search on a Boolean Circuit Synthesis Problem of Koza's 355
Abstract 355
1 Introduction 355
2 Genetic Programming 355
3 GP vs RGAT 356
4 Hill Climbing 356
5 Interpretation and Speculation 356
6 References 357
Chapter 41. Case-Based Acquisition of Place Knowledge 359
Abstract 359
1. Introduction and Basic Concepts 359
2. The Evidence Grid Representation 360
3. Case-Based Recognition of Places 361
4. Case-Based Learning of Places 362
5. Experiments with Place Learning 363
6. Related Work on Spatial Learning 365
7. Directions for Future Work 366
Acknowledgements 367
References 367
Chapter 42. Comparing Several Linear-threshold Learning Algorithms on Tasks Involving Superfluous Attributes 368
Abstract 368
1 INTRODUCTION 368
2 THE LEARNING TASKS 369
3 THE ALGORIT 369
4 DESCRIPTION OF THE PLOTS 371
5 CHECKING PROCEDURES 371
6 OBSERVATIONS 375
7 CONCLUSION 376
Chapter 43. Learning policies for partially observable environments: Scaling up 377
Abstract 377
1 INTRODUCTION 377
2 PARTIALLY OBSERVABLE MARKOV DECISION PROCESSES 378
3 SOME SOLUTION METHODS FOR POMDP's 379
4 HANDLING LARGER POMDP's: A HYBRID APPROACH 381
5 MORE ADVANCED REPRESENTATIONS 383
References 384
Chapter 44. Increasing the performance and consistency of classification trees by using the accuracy criterion at the leaves 386
Abstract 386
1 Introduction and Outline 386
2 Comparison of accuracy characteristics of split criteria 387
3 Revised Tree Growing Strategy 388
4 Empirical Results with revised strategy 389
Acknowledgements 390
References 390
Chapter 45. Efficient Learning with Virtual Threshold Gates 393
Abstract 393
1 Introduction 393
2 Preliminaries 395
3 The Winnow algorithms 395
4 Efficient On-line Learning of Simple Geometrical Objects When Dimension is Variable 396
5 Efficient On-line Learning of Simple Geometrical Objects When Dimension is Fixed 399
6 Conclusions 399
Acknowledgements 400
References 400
Chapter 46. Instance-Based Utile Distinctions for Reinforcement Learning with Hidden State 402
Abstract 402
1 INTRODUCTION 402
2 UTILE SUFFIX MEMORY 404
3 DETAILS OF THE ALGORITHM 404
4 EXPERIMENTAL RESULTS 406
5 RELATED WORK 409
6 DISCUSSION 409
Acknowledgments 410
References 410
Chapter 47. Efficient Learning from Delayed Rewards through Symbiotic Evolution 411
Abstract 411
1 Introduction 411
2 Neuro-Evolution 412
3 Symbiotic Evolution 412
4 The SANE Method 412
5 The Inverted Pendulum Problem 413
6 Population Dynamics in SANE 417
7 Related Work 417
8 Extending SANE 418
9 Conclusion 418
Acknowledgments 418
References 418
Chapter 48. Free to Choose: Investigating the Sample Complexity of Active Learning of Real Valued Functions 420
Abstract 420
1 INTRODUCTION 420
2 MODEL AND PRELIMINARIES 421
3 COLLECTING EXAMPLES: SAMPLING STRATEGIES 421
4 EXAMPLE 1: MONOTONIC FUNCTIONS 422
5 EXAMPLE 2: A CLASS WITH BOUNDED FIRST DERIVATIVE 424
6 CONCLUSIONS AND EXTENSIONS 426
Acknowledgements 426
References 426
Chapter 49. On learning Decision Committees 428
Abstract 428
1 Introduction 428
2 Definitions and theoretical results 429
3 Learning by DC{-i,0,i}: the IDC algorithm 430
4 Experiments 432
5 Discussion 433
References 434
Chapter 50. Inferring Reduced Ordered Decision Graphs of Minimum Description Length 436
Abstract 436
1 INTRODUCTION 436
2 DECISION TREES AND DECISION GRAPHS 436
3 MANIPULATING DISCRETE FUNCTIONS USING RODGS 437
4 MINIMUM MESSAGE LENGTH AND ENCODING OF RODGS 438
5 DERIVING AN RODG OF MINIMAL COMPLEXITY 439
6 EXPERIMENTS 442
7 CONCLUSIONS AND FUTURE WORK 444
References 444
Chapter 51. On Pruning and Averaging Decision Trees 445
Abstract 445
1 INTRODUCTION 445
2. OPTIMAL PRUNING 445
3 TREE AVERAGING 445
4 WEIGHTS FOR DECISION TREES 447
5 COMPLEXITY OF FANNING 448
6 COMPARISON OF AVERAGING AND PRUNING 449
7 DISCUSSION 450
8 FANNING OVER GRAPHS AND PRODUCTION RULES 451
9 CONCLUSION 451
References 452
Chapter 52. Efficient Memory-Based Dynamic Programming 453
Abstract 453
1 INTRODUCTION 453
2 MEMORY-BASED APPROACH 454
3 EXPERIMENTAL DEMONSTRATION 457
4 DISCUSSION 459
5 CONCLUSION 460
Acknowledgements 460
References 460
Chapter 53. Using Multidimensional Projection to Find Relations 462
Abstract 462
1 MOTIVATION 462
2 BASIC NOTIONS: RELATION AND PROJECTION 463
3 MULTIDIMENSIONAL RELATIONAL PROJECTION 463
4 A PROTOTYPE IMPLEMENTATION: MRP 464
5 EXPERIMENTAL RESULTS 466
6 RELATED RESEARCH 469
7 CONCLUSIONS 469
Acknowledgements 470
References 470
Chapter 54. Compression-Based Discretization of Continuous Attributes 471
Abstract 471
1 INTRODUCTION 471
2 AN MDL MEASURE FOR DISCRETIZED ATTRIBUTES 472
3 ALGORITHMIC USAGE 473
4 EXPERIMENTS AND EMPIRICAL RESULTS 474
5 CONCLUSIONS AND FURTHER RESEARCH 477
Acknowledgements 478
References 478
Chapter 55. MDL and Categorical Theories (Continued) 479
Abstract 479
1 INTRODUCTION 479
2 CLASS DESCRIPTION THEORIES AND MDL 480
3 AN ANOMALY AND A PREVIOUS SOLUTION 481
4 A NEW SOLUTION 481
5 APPLYING THE SCHEME TO C4.5RULES 482
6 RELATED RESEARCH 483
7 CONCLUSION 484
References 484
Chapter 56. For Every Generalization Action, Is There Reallyan Equal and Opposite Reaction? Analysis of the Conservation Law for Generalization Performance 486
Abstract 486
1 INTRODUCTION 486
2 CONSERVATION LAWREVISITED 486
3 AN ALTERNATE MEASURE OF GENERALIZATION 489
4 DISCUSSION 492
Acknowledgments 493
References 493
Chapter 57. Active Exploration and Learning in Real-Valued Spaces using Multi-Armed Bandit Allocation Indices 495
Abstract 495
1 Introduction and Motivation 495
2 Combining Classification Tree Algorithms with Gittins Indices 498
3 The Grasping Task 499
4 Discussion 500
5 Conclusion 501
Acknowledgments 501
References 502
Chapter 58. Discovering Solutions with Low Kolmogorov Complexity and High Generalization Capability 503
Abstract 503
1 INTRODUCTION 503
2 BASIC CONCEPTS 504
3 PROBABILISTIC SEARCH 505
4 "SIMPLE" NEURAL NETS 507
5 INCREMENTAL LEARNING 509
6 ACKNOWLEDGEMENTS 511
References 511
Chapter 59. A Comparison of Induction Algorithms for Selective andnon-Selective Bayesian Classifiers 512
Abstract 512
1 INTRODUCTION 512
2 NAIVE BAYESIAN CLASSIFIERS 513
3 BAYESIAN NETWORK CLASSIFIERS 513
5 DISCUSSION 516
6 RELATED WORK 518
7 CONCLUSION 519
Acknowledgement 520
References 520
Chapter 60. Retrofitting Decision Tree Classifiers Using Kernel Density Estimation 521
Abstract 521
1. INTRODUCTION 521
2 A REVIEW OF KERNEL DENSITY ESTIMATION 522
3 CLASSIFICATION WITH KERNEL DENSITY ESTIMATES 523
4 DECISION TREE DENSITY ESTIMATORS 523
5 DETAILS ON DECISION TREE DENSITY ESTIMATORS 524
6 EXPERIMENTAL RESULTS 524
7 RELATED WORK, EXTENSIONS, AND DISCUSSION 527
8 CONCLUSION 528
Chapter 61. Automatic Speaker Recognition: An Application of Machine Learning 530
Abstract 530
1 INTRODUCTION 530
2 PREPROCESSING 531
3 SPEAKER CLASSIFICATION 532
4 EXPERIMENTAL RESULTS 533
5 CONCLUSION 536
Acknowledgments 536
References 536
Chapter 62. An Inductive Learning Approach to Prognostic Prediction 537
Abstract 537
1 INTRODUCTION 537
2 RECURRENCE SURFACE APPROXIMATION 538
3 CLINICAL APPLICATION 542
4 CONCLUSIONS AND FUTURE WORK 544
Chapter 63. TD Models: Modeling the World at a Mixture of Time Scales 546
Abstract 546
1 Multi-Scale Planning and Modeling 546
2 Reinforcement Learning 547
3 The Prediction Problem 547
4 A Generalized Bellman Equation 548
5 n-Step Models 548
6 Intermixing Time Scales 548
7 ß-Models 549
8 Theoretical Results 550
9 TD(.) Learning of ß-models 551
10 A Wall-Following Example 551
11 A Hidden-State Example 552
12 Adding Actions (Future Work) 553
13 Conclusions 553
Acknowledgments 554
References 554
Chapter 64. Learning Collection Fusion Strategies for Information Retrieval 555
Abstract 555
1 INTRODUCTION 555
2 UNDERPINNINGS 556
3 LEARNING COLLECTION FUSION STRATEGIES 558
4 EXPERIMENTS 561
5 DISCUSSION AND CONCLUSIONS 562
References 563
Chapter 65. Learning by Observation and Practice:An Incremental Approach for Planning Operator Acquisition 564
Abstract 564
1 Introduction 564
2 Learning architecture overview 565
3 Issues of learning planning operators 565
4 Learning algorithm descriptions 567
5 Empirical results and analysis 570
Acknowledgements 571
References 572
Chapter 66. Learning with Rare Cases and Small Disjuncts 573
Abstract 573
1. INTRODUCTION 573
2. BACKGROUND 573
3. WHY ARE SMALL DISJUNCTS SO ERROR PRONE? 574
4. THE PROBLEM DOMAINS 574
5. THE EXPERIMENTS 575
6. RESULTS AND DISCUSSION 576
7. FUTURE RESEARCH 579
8. CONCLUSION 579
Acknowledgements 580
References 580
Chapter 67. Horizontal Generalization 581
Abstract 581
1 INTRODUCTION 581
2 FAN GENERALIZERS 582
3 COMPUTER EXPERIMENTS 582
4 GENERAL COMMENTS ON FG's 589
Acknowledgements 589
References 589
Chapter 68. Learning Hierarchies from Ambiguous Natural Language Data 590
Abstract 590
1 Introduction 590
2 Background 591
3 Learning Translation Rules with FOCL 591
4 Learning a Semantic Hierarchy from scratch 593
5 Updating an existing hierarchy 594
7 Limitation 597
8 Related Work 597
9 Conclusion 597
Acknowledgement 598
References 598
PART 2: INVITED TALKS 600
Chapter 69. Machine Learning and Information Retrieval 602
Chapter 70. Learning With Bayesian Networks 603
References 603
Chapter 71. Learning for Automotive Collision Avoidance and Autonomous Control 604
Author Index 606

Produkt-Details

• Erscheint lt. Verlag: 28.6.2014
• Sprache: englisch
• Themenwelt: Informatik ► Theorie / Studium ► Künstliche Intelligenz / Robotik
• ISBN-10: 1-4832-9866-3 / 1483298663
• ISBN-13: 978-1-4832-9866-5 / 9781483298665