模式识别与神经网络

1 Introduction and Examples1

1.1 How do neural methods differ?4

1.2 The patterm recognition task5

1.3 Overview of the remaining chapters9

1.4 Examples10

1.5 Literature15

2 Statistical Decision Theory17

2.1 Bayes rules for known distributions18

2.2 Parametric models26

2.3 Logistic discrimination43

2.4 Predictive classification45

2.5Alternative estimation procedures55

2.6 How complex a model do we need?59

2.7 Performance assessment66

2.8 Computational learning approaches77

3 Linear DiscriminantAnalysis91

3.1 Classical linear discriminatio92

3.2 Linear discriminants via regression101

3.3 Robustness105

3.4 Shrinkage methods106

3.5 Logistic discrimination109

3.6 Linear separatio andperceptrons116

4 Flexible Diseriminants121

4.1 Fitting smooth parametric functions122

4.2 Radial basis functions131

4.3 Regularization136

5 Feed-forward Neural Networks143

5.1 Biological motivation145

5.2 Theory147

5.3 Learning algorithms148

5.4 Examples160

5.5 Bayesian perspectives163

5.6 Network complexity168

5.7Approximation results173

6 Non-parametric Methods181

6.1 Non-parametric estlmation of class densities181

6.2 Nearest neighbour methods191

6 3 Learning vector quantization201

6.4 Mixture representations207

7 Tree-structured Classifiers213

7.1 Splitting rules216

7.2 Pruning rules221

7.3 Missing values231

7.4 Earlier approaches235

7.5 Refinements237

7.6 Relationships to neural networks240

7.7 Bayesian trees241

8 Belief Networks243

8.1 Graphical models and networks246

8.2 Causal networks262

8 3 Learning the network structure275

8.4 Boltzmann machines279

8.5 Hierarchical mixtures of experts283

9 Unsupervised Methods287

9.1 Projection methods288

9.2 Multidimensional scaling305

9.3 Clustering algorithms311

9.4 Self-organizing maps322

10 Finding Good Pattern Features327

10.1 Bounds for the Bayes error328

10.2 Normal class distributions329

10.3 Branch-and-bound techniques330

10.4 Feature extraction331

A Statistical Sidelines333

A.1 Maximum likelihood and MAP estimation333

A.2 TheEMalgorithm334

A.3 Markov chain Monte Carlo337

A.4Axioms for dconditional indcpcndence339

A.5 Oprimization342

Glossary347

References355

Author Index391

Subject Index399

《模式识别与神经网络(英文版)》是模式识别和神经网络方面的名著，讲述了模式识别所涉及的统计方法、神经网络和机器学习等分支。书的内容从介绍和例子开始，主要涵盖统计决策理论、线性判别分析、弹性判别分析、前馈神经网络、非参数方法、树结构分类、信念网、无监管方法、探寻优良的模式特性等方面的内容。
　　《模式识别与神经网络(英文版)》可作为统计与理工科研究生课程的教材，对模式识别和神经网络领域的研究人员也是极有价值的参考书。

随着人工智能、信息检索和海量数据处理等技术的发展，模式识别成为了研究热点。在《模式识别与神经网络(英文版)》中，Riprley将模式识别领域中的统计方法和基于神经网络的机器学习这两个关键思想结令起来：以统计决策理；仑和计算学习理论为依据，建立了神经网络理论的坚实基础。在理论层面，《模式识别与神经网络(英文版)》强调概率与统计；在实践层面。则强调模式识别的实用方法。
　　《模式识别与神经网络(英文版)》已被国际知名大学采用为教材，对于研究模式识别和神经网络的专业人士，也是不可不读的优秀参考书。