[ Download options ] alternative direct download for Learning Bayesian Networks(Neapolitan, Richard) pdf from usenet with usenext client 5x faster.
Usenet was created before the internet and consists of more than 60000 boards for discussions (newsgroups).
Opinions are exchanged in these boards.There is nothing you won't find there... or download torrent.
Before download check the report, the internal files and the comments of this torrent.
Download this torrent or use Magnet Link
| Add to your bookmarks |
| Learning Bayesian Networks(Neapolitan, Richard) pdf.torrent | |
| ↓ Alternative Direct Download *FREE 5x FASTER*. Click here to download the usenext client. | |
| Secure download hide your personal activity while downloading torrents with torrent privacy |
| To Download From Site You Will Need Bittorrent Software Installed.Get It Here: Visit BitRoll |
| Category : Misc » Ebooks |
| Added : 85 weeks ago |
| Size : 4.64 MB |
| Seeds : 45 |
| Peers : 4 |
| Hash : 3186babfbce53f4b200ad01bd6e2cd9a5bef37be |
| Tags : Learning Bayesian Networks |
Learning Bayesian Networks
Richard E. Neapolitan - Prentice Hall (2003) - PDF 4,57 MB
Cover book
Contents
Preface ix
I Basics 1
1 Introduction to Bayesian Networks 3
1.1 Basics of Probability Theory . . . . . . . . . . . . . . . . . . . . 5
1.1.1 Probability Functions and Spaces . . . . . . . . . . . . . . 6
1.1.2 Conditional Probability and Independence . . . . . . . . . 9
1.1.3 Bayes’ Theorem . . . . . . . . . . . . . . . . . . . . . . . 12
1.1.4 Random Variables and Joint Probability Distributions . . 13
1.2 Bayesian Inference . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.2.1 Random Variables and Probabilities in Bayesian Applications . . . . . . . 20
1.2.2 A Definition of Random Variables and Joint Probability
Distributions for Bayesian Inference . . . . . . . . . . . . 24
1.2.3 A Classical Example of Bayesian Inference . . . . . . . . . 27
1.3 Large Instances / Bayesian Networks . . . . . . . . . . . . . . . . 29
1.3.1 The Difficulties Inherent in Large Instances . . . . . . . . 29
1.3.2 TheMarkov Condition . . . . . . . . . . . . . . . . . . . . 31
1.3.3 Bayesian Networks . . . . . . . . . . . . . . . . . . . . . . 40
1.3.4 A Large BayesianNetwork . . . . . . . . . . . . . . . . . 43
1.4 Creating BayesianNetworks Using Causal Edges . . . . . . . . . 43
1.4.1 Ascertaining Causal Influences Using Manipulation . . . . 44
1.4.2 Causation and theMarkov Condition . . . . . . . . . . . 51
2 More DAG/Probability Relationships 65
2.1 Entailed Conditional Independencies . . . . . . . . . . . . . . . . 66
2.1.1 Examples of Entailed Conditional Independencies . . . . . 66
2.1.2 d-Separation . . . . . . . . . . . . . . . . . . . . . . . . . 70
2.1.3 Finding d-Separations . . . . . . . . . . . . . . . . . . . . 76
2.2 Markov Equivalence . . . . . . . . . . . . . . . . . . . . . . . . . 84
2.3 EntailingDependencies with aDAG . . . . . . . . . . . . . . . . 92
2.3.1 Faithfulness . . . . . . . . . . . . . . . . . . . . . . . . . . 95
2.3.2 Embedded Faithfulness . . . . . . . . . . . . . . . . . . . 99
2.4 Minimality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
2.5 Markov Blankets and Boundaries . . . . . . . . . . . . . . . . . . 108
2.6 More on Causal DAGs . . . . . . . . . . . . . . . . . . . . . . . . 110
2.6.1 The CausalMinimality Assumption . . . . . . . . . . . . 110
2.6.2 The Causal Faithfulness Assumption . . . . . . . . . . . . 111
2.6.3 The Causal Embedded Faithfulness Assumption . . . . . 112
II Inference 121
3 Inference: Discrete Variables 123
3.1 Examples of Inference . . . . . . . . . . . . . . . . . . . . . . . . 124
3.2 Pearl’sMessage-Passing Algorithm . . . . . . . . . . . . . . . . . 126
3.2.1 Inference in Trees . . . . . . . . . . . . . . . . . . . . . . . 127
3.2.2 Inference in Singly-Connected Networks . . . . . . . . . . 142
3.2.3 Inference inMultiply-Connected Networks . . . . . . . . . 153
3.2.4 Complexity of the Algorithm . . . . . . . . . . . . . . . . 155
3.3 The Noisy OR-GateModel . . . . . . . . . . . . . . . . . . . . . 156
3.3.1 TheModel . . . . . . . . . . . . . . . . . . . . . . . . . . 156
3.3.2 Doing InferenceWith theModel . . . . . . . . . . . . . . 160
3.3.3 FurtherModels . . . . . . . . . . . . . . . . . . . . . . . . 161
3.4 Other Algorithms that Employ the DAG. . . . . . . . . . . . . . 161
3.5 The SPI Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . 162
3.5.1 The Optimal Factoring Problem . . . . . . . . . . . . . . 163
3.5.2 Application to Probabilistic Inference . . . . . . . . . . . 168
3.6 Complexity of Inference . . . . . . . . . . . . . . . . . . . . . . . 170
3.7 Relationship to Human Reasoning . . . . . . . . . . . . . . . . . 171
3.7.1 The Causal NetworkModel . . . . . . . . . . . . . . . . . 171
3.7.2 Studies Testing the Causal NetworkModel . . . . . . . . 173
4 More Inference Algorithms 181
4.1 Continuous Variable Inference . . . . . . . . . . . . . . . . . . . . 181
4.1.1 The Normal Distribution . . . . . . . . . . . . . . . . . . 182
4.1.2 An Example Concerning Continuous Variables . . . . . . 183
4.1.3 An Algorithm for Continuous Variables . . . . . . . . . . 185
4.2 Approximate Inference . . . . . . . . . . . . . . . . . . . . . . . . 205
4.2.1 A Brief Review of Sampling . . . . . . . . . . . . . . . . . 205
4.2.2 Logic Sampling . . . . . . . . . . . . . . . . . . . . . . . . 211
4.2.3 LikelihoodWeighting . . . . . . . . . . . . . . . . . . . . . 217
4.3 Abductive Inference . . . . . . . . . . . . . . . . . . . . . . . . . 221
4.3.1 Abductive Inference in Bayesian Networks . . . . . . . . . 221
4.3.2 A Best-First Search Algorithm for Abductive Inference . . 224
5 Influence Diagrams 239
5.1 Decision Trees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
5.1.1 Simple Examples . . . . . . . . . . . . . . . . . . . . . . . 239
5.1.2 Probabilities, Time, and Risk Attitudes . . . . . . . . . . 242
5.1.3 SolvingDecision Trees . . . . . . . . . . . . . . . . . . . . 245
5.1.4 More Examples . . . . . . . . . . . . . . . . . . . . . . . . 245
5.2 Influence Diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . 259
5.2.1 Representing with Influence Diagrams . . . . . . . . . . . 259
5.2.2 Solving Influence Diagrams . . . . . . . . . . . . . . . . . 266
5.3 Dynamic Networks . . . . . . . . . . . . . . . . . . . . . . . . . . 272
5.3.1 Dynamic Bayesian Networks . . . . . . . . . . . . . . . . 272
5.3.2 Dynamic Influence Diagrams . . . . . . . . . . . . . . . . 279
III Learning 291
6 Parameter Learning: Binary Variables 293
6.1 Learning a Single Parameter . . . . . . . . . . . . . . . . . . . . . 294
6.1.1 Probability Distributions of Relative Frequencies . . . . . 294
6.1.2 Learning a Relative Frequency . . . . . . . . . . . . . . . 303
6.2 More on the Beta Density Function . . . . . . . . . . . . . . . . . 310
6.2.1 Non-integral Values of a and b . . . . . . . . . . . . . . . 311
6.2.2 Assessing the Values of a and b . . . . . . . . . . . . . . . 313
6.2.3 Why the BetaDensity Function? . . . . . . . . . . . . . . 315
6.3 Computing a Probability Interval . . . . . . . . . . . . . . . . . . 319
6.4 Learning Parameters in a Bayesian Network . . . . . . . . . . . . 323
6.4.1 Urn Examples . . . . . . . . . . . . . . . . . . . . . . . . 323
6.4.2 Augmented Bayesian Networks . . . . . . . . . . . . . . . 331
6.4.3 Learning Using an Augmented Bayesian Network . . . . . 336
6.4.4 A Problem with Updating; Using an Equivalent Sample
Size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 348
6.5 Learning withMissing Data Items . . . . . . . . . . . . . . . . . 357
6.5.1 Data ItemsMissing at Random . . . . . . . . . . . . . . . 358
6.5.2 Data ItemsMissing Not at Random . . . . . . . . . . . . 363
6.6 Variances in ComputedRelative Frequencies . . . . . . . . . . . . 364
6.6.1 A Simple Variance Determination . . . . . . . . . . . . . 364
6.6.2 The Variance and Equivalent Sample Size . . . . . . . . . 366
6.6.3 Computing Variances in Larger Networks . . . . . . . . . 372
6.6.4 When Do Variances Become Large? . . . . . . . . . . . . 373
7 More Parameter Learning 381
7.1 Multinomial Variables . . . . . . . . . . . . . . . . . . . . . . . . 381
7.1.1 Learning a Single Parameter . . . . . . . . . . . . . . . . 381
7.1.2 More on the Dirichlet Density Function . . . . . . . . . . 388
7.1.3 Computing Probability Intervals andRegions . . . . . . . 389
7.1.4 Learning Parameters in a Bayesian Network . . . . . . . . 392
7.1.5 Learning withMissing Data Items . . . . . . . . . . . . . 398
7.1.6 Variances in Computed Relative Frequencies . . . . . . . 398
7.2 Continuous Variables . . . . . . . . . . . . . . . . . . . . . . . . . 398
7.2.1 Normally Distributed Variable . . . . . . . . . . . . . . . 399
7.2.2 Multivariate Normally Distributed Variables . . . . . . . 413
7.2.3 Gaussian Bayesian Networks . . . . . . . . . . . . . . . . 425
8 Bayesian Structure Learning 441
8.1 Learning Structure: Discrete Variables . . . . . . . . . . . . . . . 441
8.1.1 Schema for Learning Structure . . . . . . . . . . . . . . . 442
8.1.2 Procedure for Learning Structure . . . . . . . . . . . . . . 445
8.1.3 Learning From a Mixture of Observational and Experimental
Data. . . . . . . . . . . . . . . . . . . . . . . . . . 449
8.1.4 Complexity of Structure Learning . . . . . . . . . . . . . 450
8.2 Model Averaging . . . . . . . . . . . . . . . . . . . . . . . . . . . 451
8.3 Learning Structure withMissingData . . . . . . . . . . . . . . . 452
8.3.1 Monte CarloMethods . . . . . . . . . . . . . . . . . . . . 453
8.3.2 Large-Sample Approximations . . . . . . . . . . . . . . . 462
8.4 ProbabilisticModel Selection . . . . . . . . . . . . . . . . . . . . 468
8.4.1 ProbabilisticModels . . . . . . . . . . . . . . . . . . . . . 468
8.4.2 TheModel Selection Problem . . . . . . . . . . . . . . . . 472
8.4.3 Using the Bayesian Scoring Criterion for Model Selection 473
8.5 Hidden Variable DAGModels . . . . . . . . . . . . . . . . . . . . 476
8.5.1 Models ContainingMore Conditional Independencies than
DAGModels . . . . . . . . . . . . . . . . . . . . . . . . . 477
8.5.2 Models Containing the Same Conditional Independencies
as DAGModels . . . . . . . . . . . . . . . . . . . . . . . . 479
8.5.3 Dimension of Hidden Variable DAGModels . . . . . . . . 484
8.5.4 Number ofModels andHidden Variables . . . . . . . . . . 486
8.5.5 EfficientModel Scoring . . . . . . . . . . . . . . . . . . . 487
8.6 Learning Structure: Continuous Variables . . . . . . . . . . . . . 491
8.6.1 The Density Function of D . . . . . . . . . . . . . . . . . 491
8.6.2 The Density function of D Given aDAGpattern . . . . . 495
8.7 Learning Dynamic Bayesian Networks . . . . . . . . . . . . . . . 505
9 Approximate Bayesian Structure Learning 511
9.1 ApproximateModel Selection . . . . . . . . . . . . . . . . . . . . 511
9.1.1 Algorithms that Search over DAGs . . . . . . . . . . . . . 513
9.1.2 Algorithms that Search over DAGPatterns . . . . . . . . 518
9.1.3 An Algorithm Assuming Missing Data or Hidden Variables529
9.2 ApproximateModel Averaging . . . . . . . . . . . . . . . . . . . 531
9.2.1 AModel Averaging Example . . . . . . . . . . . . . . . . 532
9.2.2 ApproximateModel Averaging UsingMCMC . . . . . . . 533
10 Constraint-Based Learning 541
10.1 Algorithms Assuming Faithfulness . . . . . . . . . . . . . . . . . 542
10.1.1 Simple Examples . . . . . . . . . . . . . . . . . . . . . . . 542
10.1.2 Algorithms for Determining DAGpatterns . . . . . . . . 545
10.1.3 Determining if a Set Admits a Faithful DAG Representation552
10.1.4 Application to Probability . . . . . . . . . . . . . . . . . . 560
10.2 Assuming Only Embedded Faithfulness . . . . . . . . . . . . . . 561
10.2.1 Inducing Chains . . . . . . . . . . . . . . . . . . . . . . . 562
10.2.2 A Basic Algorithm . . . . . . . . . . . . . . . . . . . . . . 568
10.2.3 Application to Probability . . . . . . . . . . . . . . . . . . 590
10.2.4 Application to Learning Causal Influences1 . . . . . . . . 591
10.3 Obtaining the d-separations . . . . . . . . . . . . . . . . . . . . . 599
10.3.1 Discrete Bayesian Networks . . . . . . . . . . . . . . . . . 600
10.3.2 Gaussian Bayesian Networks . . . . . . . . . . . . . . . . 603
10.4 Relationship to Human Reasoning . . . . . . . . . . . . . . . . . 604
10.4.1 Background Theory . . . . . . . . . . . . . . . . . . . . . 604
10.4.2 A Statistical Notion of Causality . . . . . . . . . . . . . . 606
11 More Structure Learning 617
11.1 Comparing theMethods . . . . . . . . . . . . . . . . . . . . . . . 617
11.1.1 A Simple Example . . . . . . . . . . . . . . . . . . . . . . 618
11.1.2 Learning College Attendance Influences . . . . . . . . . . 620
11.1.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . 623
11.2 Data Compression Scoring Criteria . . . . . . . . . . . . . . . . . 624
11.3 Parallel Learning of Bayesian Networks . . . . . . . . . . . . . . 624
11.4 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 624
11.4.1 Structure Learning . . . . . . . . . . . . . . . . . . . . . . 625
11.4.2 Inferring Causal Relationships . . . . . . . . . . . . . . . 633
IV Applications 647
12 Applications 649
12.1 Applications Based on Bayesian Networks . . . . . . . . . . . . . 649
12.2 Beyond Bayesian networks . . . . . . . . . . . . . . . . . . . . . . 655
Bibliography 657
Index 686
