Supervised and unsupervised pattern recognition : feature extraction and computational intelligence / Evangelia Micheli-Tzanako.

Format:

Book

Author/Creator:

Micheli-Tzanakou, Evangelia, 1942-

Contributor:

Class of 1932 Fund.

Series:

Industrial electronics series

Language:

English

Subjects (All):

Pattern recognition systems.

Neural networks (Computer science).

Physical Description:

371 pages : illustrations ; 25 cm.

Place of Publication:

Boca Raton, Fla. : CRC Press, [2000]

Summary:

Describes the application of supervised and unsupervised pattern recognition schemes to the classification of various types of waveforms and images. Micheli-Tzanakou (biomedical engineering, (Rutgers U.) treats experimental and theoretical contributions equally and examines interchanges between the two. The chapters span a variety of problems in signal and image processing, using mainly neural networks for classification and template matching.

Contents:

Section I Overviews of Neural Networks, Classifiers, and Feature Extraction Methods - Supervised Neural Networks

1.2 Criteria for Optimal Classifier Design 3

1.3 Categorizing the Classifiers 4

1.3.1 Bayesian Optimal Classifiers 4

1.3.2 Exemplar Classifiers 5

1.3.3 Space Partition Methods 6

1.3.4 Neural Networks 7

1.4.1 Bayesian Classifiers 7

1.4.1.1 Minimum ECM Classifers 8

1.4.1.2 Multi-Class Optimal Classifiers 9

1.4.2 Bayesian Classifiers with Multivariate Normal Populations 11

1.4.2.1 Quadratic Discriminant Score 11

1.4.2.2 Linear Discriminant Score 11

1.4.2.3 Linear Discriminant Analysis and Classification 12

1.4.2.4 Equivalence of LDF to Minimum TPM Classifier 14

1.4.3 Learning Vector Quantizer (LVQ) 14

1.4.3.1 Competitive Learning 14

1.4.3.2 Self-Organizing Map 15

1.4.3.3 Learning Vector Quantization 15

1.4.4 Nearest Neighbor Rule 18

1.5 Neural Networks (NN) 19

1.5.1.1 Artificial Neural Networks 19

1.5.1.2 Usage of Neural Networks 19

1.5.1.3 Other Neural Networks 20

1.5.2 Feed-Forward Neural Networks 20

1.5.3 Error Backpropagation 22

1.5.3.1 Madaline Rule III for Multilayer Network with Sigmoid Function 25

1.5.3.2 A Comment on the Terminology 'Backpropagation' 25

1.5.3.3 Optimization Machines with Feed-Forward Multilayer Perceptrons 25

1.5.3.4 Justification for Gradient Methods for Nonlinear Function Approximation 26

1.5.3.5 Training Methods for Feed-Forward Networks 27

1.5.4 Issues in Neural Networks 28

1.5.4.1 Universal Approximation 28

1.5.5 Enhancing Convergence Rate and Generalization of an Optimization Machine 29

1.5.5.1 Suggestions for Improving the Convergence 30

1.5.5.2 Quick Prop 31

1.5.5.3 Kullback-Leibler Distance 32

1.5.5.4 Weight Decay 33

1.5.5.5 Regression Methods for Classification Purposes 34

1.5.6 Two-Group Regression and Linear Discriminant Function 34

1.5.7 Multi-Response Regression and Flexible Discriminant Analysis 36

1.5.7.1 Powerful Nonparametric Regression Methods for Classification Problems 37

1.5.8 Optimal Scoring (OS) 37

1.5.8.1 Partially Minimized ASR 39

1.5.9 Canonical Correlation Analysis 40

1.5.10 Linear Discriminant Analysis 41

1.5.10.1 LDA Revisited 41

1.5.11 Translation of Optimal Scoring Dimensions into Discriminant Coordinates 42

1.5.12 Linear Discriminant Analysis via Optimal Scoring 44

1.5.12.1 LDA via OS 45

1.5.13 Flexible Discriminant Analysis by Optimal Scoring 46

1.6 Comparison of Experimental Results 48

1.7 System Performance Assessment 49

1.7.1 Classifier Evaluation 50

1.7.1.1 Hold-Out Method 51

1.7.1.2 K-Fold Cross-Validation 51

1.7.2 Bootstrapping Method for Estimation 52

1.7.2.1 Jackknife Estimation 53

1.7.2.2 Bootstrap Method 54

1.8 Analysis of Prediction Rates from Bootstrapping Assessment 54

Chapter 2 Artificial Neural Networks: Definitions, Methods, Applications

2.3 Training Algorithms 64

2.3.1 Backpropagation Algorithm 65

2.3.2 The ALOPEX Algorithm 69

2.3.3 Multilayer Perceptron (MLP) Network Training with ALOPEX 71

2.4 Some Applications 72

2.4.1 Expert Systems and Neural Networks 72

2.4.2 Applications in Mammography 73

2.4.3 Chromosome and Genetic Sequences Classification 74

Chapter 3 A System for Handwritten Digit Recognition

3.2 Preprocessing of Handwritten Digit Images 79

3.2.1 Optimal Size of the Mask for Dilation 85

3.2.2 Bartlett Statistic 85

3.3 Zernike Moments (ZM) for Characterization of Image Patterns 87

3.3.1 Reconstruction by Zernike Moments 90

3.3.2 Features from Zernike Moments 92

3.4 Dimensionality Reduction 96

3.4.1 Principal Component Analysis 96

3.4.2 Discriminant Analysis 98

3.5 Analysis of Prediction Error Rates from Bootstrapping Assessment 100

Chapter 4 Other Types of Feature Extraction Methods

4.2 Wavelets 110

4.2.1 Discrete Wavelet Series 111

4.2.2 Discrete Wavelet Transform (DWT) 112

4.2.3 Spline Wavelet Transform 112

4.2.4 The Discrete B-Spline Wavelet Transform 114

4.2.5 Design of Quadratic Spline Wavelets 114

4.2.6 The Fast Algorithm 117

4.3 Invariant Moments 119

4.4 Entropy 122

4.5 Cepstrum Analysis 122

4.6 Fractal Dimension 123

4.7 SGLD Texture Features 126

Section II Unsupervised Neural Networks

Chapter 5 Fuzzy Neural Networks

5.2 Pattern Recognition 135

5.2.1 Theory and Applications 135

5.2.2 Feature Extraction 137

5.2.3 Clustering 138

5.3 Optimization 138

5.3.1 Theory and Objectives 138

5.3.3 Modified ALOPEX Algorithm 141

5.4 System Design 144

5.4.1 Feature Extraction 144

5.4.1.1 The Karhunen-Loeve Expansion 145

5.4.1.2 Application by a Neural Network 147

5.5 Clustering 153

5.5.1 The Fuzzy c-Means (FCM) Clustering Algorithm 153

Chapter 6 Application to Handwritten Digits

6.1 Introduction to Character Recognition 163

6.2 Data Collection 165

6.2.1 Preprocessing 166

6.2.2 Noise Thresholding 166

6.2.3 Center of Mass Adjustment 168

6.2.4 Line Thinning 168

6.2.5 Fixing to Size 168

6.2.6 Rotation 168

6.2.7 Reducing Resolution 169

6.2.8 Blurring 170

Chapter 7 An Unsupervised Neural Network System for Visual Evoked Potentials

7.2 Data Collection and Preprocessing 186

7.3 System Design 187

Section III Advanced Neural Network Architectures/Modular Neural Networks

Chapter 8 Classification of Mammograms Using a Modular Neural Network

8.2 Methods and System Overview 199

8.2.1 Data Acquisition 199

8.2.2 Feature Extraction by Transformation 200

8.3 Modular Neural Networks 202

8.4 Neural Network Training 203

8.5 Classification Results 203

8.6 The Process of Obtaining Results 207

8.7 ALOPEX Parameters 209

Chapter 9 Visual Ophthalmologist: An Automated System for Classification of Retinal Damage

9.2.1 Image Processing 223

9.2.2 Feature Extraction Methods 223

9.2.3 Image Classification 223

9.3 Modular Neural Networks 223

9.4 Application to Ophthalmology 224

Chapter 10 A Three-Dimensional Neural Network Architecture

10.2 The Neural Network Architecture 229

10.3 Simulations 230

10.3.1 Visual Receptive Fields 231

10.3.2 Modeling of Parkinson's Disease 235

Section IV General Applications

Chapter 11 A Feature Extraction Algorithm Using Connectivity Strengths and Moment Invariants

11.2 ALOPEX Algorithms 242

11.2.1 Original Algorithm 242

11.2.2 Reinforcement Rules 242

11.2.3 A Generalized ALOPEX Algorithm 243

11.3 Moment Invariants and ALOPEX 246

Chapter 12 Multilayer Perceptrons with ALOPEX: 2D-Template Matching and VLSI Implementation

12.1.1 Multilayer Perceptrons 265

12.2 Multilayer Perceptron and Template Matching 268

12.3 VLSI Implementation of ALOPEX 270

Chapter 13 Implementing Neural Networks in Silicon

13.2 The Living Neuron 278

13.3 Neuromorphic Models 280

13.4 Neurological Process Modeling 292

Chapter 14 Speaker Identification through Wavelet Multiresolution Decomposition and ALOPEX

14.2 Multiresolution Analysis through Wavelet Decomposition 303

14.3 Pattern Recognition with ALOPEX 306

14.4 Methods 306

14.4.1 Data Acquisition 306

14.4.2 Data Preprocessing 307

14.4.3 Representing the Wavelet Coefficients for Template Matching 308

Chapter 15 Face Recognition in Alzheimer's Disease: A Simulation

Chapter 16 Self-Learning Layered Neural Network

16.2 Neocognitron and Pattern Classification 325

16.2.1 Training Algorithm 328

16.5.1 Network Description 330

Chapter 17 Biological and Machine Vision

17.2 Distributed Representation 347

17.3 The Model 348

17.4 A Modified ALOPEX Algorithm 348

17.5 Application to Template Matching 350

17.6 Brain to Computer Link 351

17.6.1 Global Receptive Fields in the Human Visual System 351

17.6.2 The Black Box Approach 353.

Notes:

Includes bibliographical references and index.

Local Notes:

Acquired for the Penn Libraries with assistance from the Class of 1932 Fund.

ISBN:

0849322782

OCLC:

42080284