Bayesian artificial intelligence

Title:

Personal Author:

Korb, Kevin B.

Series:

Series in computer science and data analysis

Publication Information:

Boca Raton : Chapman & Hall/CRC, 2004

ISBN:

9781584883876

Subject Term:

Bayesian statistical decision theory -- Data processing

Machine learning

Artificial neural networks

Added Author:

Nicholson, Ann E.

Available:*

Library	Item Barcode	Call Number	Material Type	Item Category 1	Status
Searching... PSZ JB	30000010064562	QA279.5 K67 2004	Open Access Book	Book	Searching... Unknown
Searching... PSZ KL	30000010138982	QA279.5 K67 2004	Open Access Book	Book	Searching... Unknown

As the power of Bayesian techniques has become more fully realized, the field of artificial intelligence has embraced Bayesian methodology and integrated it to the point where an introduction to Bayesian techniques is now a core course in many computer science programs. Unlike other books on the subject, Bayesian Artificial Intelligencekeeps mathematical detail to a minimum and covers a broad range of topics. The authors integrate all of Bayesian net technology and learning Bayesian net technology and apply them both to knowledge engineering. They emphasize understanding and intuition but also provide the algorithms and technical background needed for applications. Software, exercises, and solutions are available on the authors' website.

Author Notes

Ann Nicholson is a Senior Lecturer and Kevin Korb is a Reader in the School of Computer Science and Software Engineering at Monash University, Victoria, Australia

Reviews 1

Choice Review

Korb and Nicholson (both, Monash Univ., Australia) say in their preface that this book is aimed at advanced undergraduates in computer science who have some background in artificial intelligence, and at those who wish to engage in applied or pure research in applications of Bayesian inference in AI. They also explain how this book is different--an emphasis on causal discovery and interpretation of Bayesian networks and discussion of applications. The book has the following noteworthy features: clear and helpful introductions and summaries for each chapter, problems for each chapter for reinforcement, lists of up-to-date references, and an annotated list of relevant software showing features availability, etc. The three parts of the book, dealing respectively with fundamental background material on probabilistic reasoning, causal models, and knowledge engineering, treat a wide span of relevant material and case studies, written lucidly and carrying readers smoothly from the simple to the complex. This book certainly deserves to be in the library of any institution where undergraduate or graduate courses in computer science are taught, and would also be an excellent resource for anyone who wants to learn more about this cutting-edge area of computing. ^BSumming Up: Essential. General readers; upper-division undergraduates through professionals; two-year technical program students. R. Bharath emeritus, Northern Michigan University

Part I Probabilistic Reasoning	p. 1
Chapter 1 Bayesian Reasoning	p. 3
1.1 Reasoning under uncertainty	p. 3
1.2 Uncertainty in AI	p. 4
1.3 Probability calculus	p. 5
1.3.1 Conditional probability theorems	p. 8
1.3.2 Variables	p. 9
1.4 Interpretations of probability	p. 10
1.5 Bayesian philosophy	p. 12
1.5.1 Bayes' theorem	p. 12
1.5.2 Betting and odds	p. 13
1.5.3 Expected utility	p. 15
1.5.4 Dutch books	p. 16
1.5.5 Bayesian reasoning examples	p. 17
1.6 The goal of Bayesian AI	p. 21
1.7 Achieving Bayesian AI	p. 22
1.8 Are Bayesian networks Bayesian?	p. 22
1.9 Summary	p. 23
1.10 Bibliographic notes	p. 23
1.11 Technical notes	p. 24
1.12 Problems	p. 25
Chapter 2 Introducing Bayesian Networks	p. 29
2.1 Introduction	p. 29
2.2 Bayesian network basics	p. 29
2.2.1 Nodes and values	p. 30
2.2.2 Structure	p. 31
2.2.3 Conditional probabilities	p. 32
2.2.4 The Markov property	p. 33
2.3 Reasoning with Bayesian networks	p. 33
2.3.1 Types of reasoning	p. 34
2.3.2 Types of evidence	p. 35
2.3.3 Reasoning with numbers	p. 36
2.4 Understanding Bayesian networks	p. 37
2.4.1 Representing the joint probability distribution	p. 37
2.4.2 Pearl's network construction algorithm	p. 37
2.4.3 Compactness and node ordering	p. 38
2.4.4 Conditional independence	p. 39
2.4.5 d-separation	p. 41
2.5 More examples	p. 43
2.5.1 Earthquake	p. 43
2.5.2 Metastatic cancer	p. 43
2.5.3 Asia	p. 43
2.6 Summary	p. 44
2.7 Bibliographic notes	p. 45
2.8 Problems	p. 47
Chapter 3 Inference in Bayesian Networks	p. 53
3.1 Introduction	p. 53
3.2 Exact inference in chains	p. 54
3.2.1 Two node network	p. 54
3.2.2 Three node chain	p. 55
3.3 Exact inference in polytrees	p. 56
3.3.1 Kim and Pearl's message passing algorithm	p. 57
3.3.2 Message passing example	p. 60
3.3.3 Algorithm features	p. 61
3.4 Inference with uncertain evidence	p. 62
3.4.1 Using a virtual node	p. 63
3.4.2 Virtual nodes in the message passing algorithm	p. 65
3.5 Exact inference in multiply-connected networks	p. 66
3.5.1 Clustering methods	p. 66
3.5.2 Junction trees	p. 68
3.6 Approximate inference with stochastic simulation	p. 72
3.6.1 Logic sampling	p. 72
3.6.2 Likelihood weighting	p. 74
3.6.3 Markov Chain Monte Carlo (MCMC)	p. 75
3.6.4 Using virtual evidence	p. 75
3.6.5 Assessing approximate inference algorithms	p. 76
3.7 Other computations	p. 77
3.7.1 Belief revision	p. 77
3.7.2 Probability of evidence	p. 78
3.8 Causal inference	p. 79
3.9 Summary	p. 81
3.10 Bibliographic notes	p. 81
3.11 Problems	p. 82
Chapter 4 Decision Networks	p. 89
4.1 Introduction	p. 89
4.2 Utilities	p. 89
4.3 Decision network basics	p. 91
4.3.1 Node types	p. 91
4.3.2 Football team example	p. 92
4.3.3 Evaluating decision networks	p. 93
4.3.4 Information links	p. 94
4.3.5 Fever example	p. 96
4.3.6 Types of actions	p. 96
4.4 Sequential decision making	p. 98
4.4.1 Test-action combination	p. 98
4.4.2 Real estate investment example	p. 99
4.4.3 Evaluation using a decision tree model	p. 101
4.4.4 Value of information	p. 103
4.4.5 Direct evaluation of decision networks	p. 104
4.5 Dynamic Bayesian networks	p. 104
4.5.1 Nodes, structure and CPTs	p. 105
4.5.2 Reasoning	p. 107
4.5.3 Inference algorithms for DBNs	p. 109
4.6 Dynamic decision networks	p. 110
4.6.1 Mobile robot example	p. 111
4.7 Summary	p. 112
4.8 Bibliographic notes	p. 113
4.9 Problems	p. 114
Chapter 5 Applications of Bayesian Networks	p. 117
5.1 Introduction	p. 117
5.2 A brief survey of BN applications	p. 118
5.2.1 Types of reasoning	p. 118
5.2.2 BN structures for medical problems	p. 118
5.2.3 Other medical applications	p. 120
5.2.4 Non-medical applications	p. 121
5.3 Bayesian poker	p. 122
5.3.1 Five-card stud poker	p. 123
5.3.2 A decision network for poker	p. 124
5.3.3 Betting with randomization	p. 127
5.3.4 Bluffing	p. 128
5.3.5 Experimental evaluation	p. 129
5.4 Ambulation monitoring and fall detection	p. 129
5.4.1 The domain	p. 129
5.4.2 The DBN model	p. 130
5.4.3 Case-based evaluation	p. 133
5.4.4 An extended sensor model	p. 134
5.5 A Nice Argument Generator (NAG)	p. 136
5.5.1 NAG architecture	p. 137
5.5.2 Example: An asteroid strike	p. 138
5.5.3 The psychology of inference	p. 139
5.5.4 Example: The asteroid strike continues	p. 141
5.5.5 The future of argumentation	p. 142
5.6 Summary	p. 142
5.7 Bibliographic notes	p. 143
5.8 Problems	p. 144
Part II Learning Causal Models	p. 147
Chapter 6 Learning Linear Causal Models	p. 151
6.1 Introduction	p. 151
6.2 Path models	p. 153
6.2.1 Wright's first decomposition rule	p. 155
6.2.2 Parameterizing linear models	p. 159
6.2.3 Learning linear models is complex	p. 159
6.3 Conditional independence learners	p. 161
6.3.1 Markov equivalence	p. 164
6.3.2 PC algorithm	p. 167
6.3.3 Causal discovery versus regression	p. 169
6.4 Summary	p. 170
6.5 Bibliographic notes	p. 170
6.6 Technical notes	p. 171
6.7 Problems	p. 172
Chapter 7 Learning Probabilities	p. 175
7.1 Introduction	p. 175
7.2 Parameterizing discrete models	p. 176
7.2.1 Parameterizing a binomial model	p. 176
7.2.2 Parameterizing a multinomial model	p. 179
7.3 Incomplete data	p. 181
7.3.1 The Bayesian solution	p. 182
7.3.2 Approximate solutions	p. 182
7.3.3 Incomplete data: summary	p. 187
7.4 Learning local structure	p. 187
7.4.1 Causal interaction	p. 187
7.4.2 Noisy-or connections	p. 188
7.4.3 Classification trees and graphs	p. 189
7.4.4 Logit models	p. 191
7.4.5 Dual model discovery	p. 191
7.5 Summary	p. 192
7.6 Bibliographic notes	p. 192
7.7 Technical notes	p. 193
7.8 Problems	p. 194
Chapter 8 Learning Discrete Causal Structure	p. 197
8.1 Introduction	p. 197
8.2 Cooper & Herskovits' K2	p. 198
8.2.1 Learning variable order	p. 200
8.3 MDL causal discovery	p. 201
8.3.1 Lam and Bacchus's MDL code for causal models	p. 203
8.3.2 Suzuki's MDL code for causal discovery	p. 205
8.4 Metric pattern discovery	p. 205
8.5 CaMML: Causal discovery via MML	p. 207
8.5.1 An MML code for causal structures	p. 207
8.5.2 An MML metric for linear models	p. 210
8.6 CaMML stochastic search	p. 211
8.6.1 Genetic algorithm (GA) search	p. 211
8.6.2 Metropolis search	p. 211
8.6.3 Prior constraints	p. 213
8.6.4 MML models	p. 214
8.6.5 An MML metric for discrete models	p. 215
8.7 Experimental evaluation	p. 215
8.7.1 Qualitative evaluation	p. 216
8.7.2 Quantiative evaluation	p. 216
8.8 Summary	p. 217
8.9 Bibliographic notes	p. 218
8.10 Technical notes	p. 218
8.11 Problems	p. 219
Part III Knowledge Engineering	p. 221
Chapter 9 Knowledge Engineering with Bayesian Networks	p. 225
9.1 Introduction	p. 225
9.1.1 Bayesian network modeling tasks	p. 225
9.2 The KEBN process	p. 226
9.2.1 KEBN lifecycle model	p. 226
9.2.2 Prototyping and spiral KEBN	p. 227
9.2.3 Are BNs suitable for the domain problem?	p. 228
9.2.4 Process management	p. 229
9.3 Modeling and elicitation	p. 230
9.3.1 Variables and values	p. 230
9.3.2 Graphical structure	p. 233
9.3.3 Probabilities	p. 241
9.3.4 Local structure	p. 247
9.3.5 Variants of Bayesian networks	p. 250
9.3.6 Modeling example: missing car	p. 251
9.3.7 Decision networks	p. 254
9.4 Adaptation	p. 257
9.4.1 Adapting parameters	p. 258
9.4.2 Structural adaptation	p. 259
9.5 Summary	p. 260
9.6 Bibliographic notes	p. 260
9.7 Problems	p. 261
Chapter 10 Evaluation	p. 263
10.1 Introduction	p. 263
10.2 Elicitation review	p. 263
10.3 Sensitivity analysis	p. 264
10.3.1 Sensitivity to evidence	p. 264
10.3.2 Sensitivity to changes in parameters	p. 271
10.4 Case-based evaluation	p. 272
10.4.1 Explanation methods	p. 273
10.5 Validation methods	p. 274
10.5.1 Predictive accuracy	p. 276
10.5.2 Expected value	p. 277
10.5.3 Kullback-Leibler divergence	p. 278
10.5.4 Information reward	p. 280
10.5.5 Bayesian information reward	p. 281
10.6 Summary	p. 283
10.7 Bibliographic notes	p. 284
10.8 Technical notes	p. 284
10.9 Problems	p. 286
Chapter 11 KEBN Case Studies	p. 287
11.1 Introduction	p. 287
11.2 Bayesian poker revisited	p. 287
11.2.1 The initial prototype	p. 287
11.2.2 Subsequent developments	p. 288
11.2.3 Ongoing Bayesian poker	p. 289
11.2.4 KEBN aspects	p. 290
11.3 An intelligent tutoring system for decimal understanding	p. 290
11.3.1 The ITS domain	p. 291
11.3.2 ITS system architecture	p. 293
11.3.3 Expert elicitation	p. 294
11.3.4 Automated methods	p. 301
11.3.5 Field trial evaluation	p. 303
11.3.6 KEBN aspects	p. 304
11.4 Seabreeze prediction	p. 305
11.4.1 The seabreeze prediction problem	p. 305
11.4.2 The data	p. 306
11.4.3 Bayesian network modeling	p. 307
11.4.4 Experimental evaluation	p. 308
11.4.5 KEBN aspects	p. 312
11.5 Summary	p. 313
Appendix A Notation	p. 315
Appendix B Software Packages	p. 317
B.1 Introduction	p. 317
B.2 History	p. 318
B.3 Murphy's Software Package Survey	p. 318
B.4 BN software	p. 323
B.4.1 Analytica	p. 323
B.4.2 BayesiaLab	p. 324
B.4.3 Bayes Net Toolbox (BNT)	p. 325
B.4.4 GeNIe	p. 326
B.4.5 Hugin	p. 327
B.4.6 JavaBayes	p. 328
B.4.7 MSBNx	p. 329
B.4.8 Netica	p. 329
B.5 Bayesian statistical modeling	p. 330
B.5.1 BUGS	p. 330
B.5.2 First Bayes	p. 331
B.6 Causal discovery programs	p. 331
B.6.1 Bayesware Discoverer	p. 331
B.6.2 CaMML	p. 331
B.6.3 TETRAD	p. 332
B.6.4 WinMine	p. 332
References	p. 333
Index	p. 355

Available:*

On Order

Summary

Summary

Author Notes

Reviews 1

Choice Review

Table of Contents