神经元网络系统设计方法 英文PDF电子书下载

1 INTRODUCTION 1

1.1 ANNs for Pattern Recognition 1

1.2 Neural Network Model 3

1.3 ANN Architecture 6

1.4 Hardware Implementation of ANN 7

1.5 A VLSI System Design Methodology 10

PARTⅠPARALLEL ANN MODELS 13

2 AN UNSUPERVISED LEARNING MODEL 13

2.1 Introduction 13

2.2 Fuzzy Clustering Neural Networks 17

2.3 Parallel FCNN Architecture 20

2.4 Experimental Results 21

2.5 Summary 24

3 A SUPERVISED TRAINING MODEL 29

3.1 Introduction 29

3.2 Linear Separability Analysis 31

3.3 Layer Adaptation Approach 35

3.4 Experiment:Pattern Recognition 39

3.5 Comparisons 41

3.6 Summary 42

4 A NEURAL-LIKE NETWORK MODEL 45

4.1 Introduction 45

4.2 FRNN Computing Model 49

4.3 FRNN Architecture 51

4.4 Case Studies 52

4.5 Summary 54

PART Ⅱ VLSI ARCHITECTURES 57

5 A PARALLEL ARCHITECTURE IMPLEMENTED BYSYSTOLIC ARRAYS 57

5.1 Introduction 57

5.2 FCNN Architecture 58

5.3 Mapping Policies 63

5.4 Typical SA Structures 65

5.5 Mapping FCNN onto SA 68

5.6 Summary 70

6 A PIPELINED ARCHITECTURE BASED ON WINDOW OPERATION 73

6.1 Introduction 73

6.2 Pipelined Architecture:Window Operation 73

6.3 Window Implementation 80

6.4 Case Studies 84

6.5 Performance Analysis 89

6.6 Summary 92

7 A SIMPLIFIED ARCHITECTURE USING A PRIORI KNOWLEDGE 93

7.1 Introduction 93

7.2 Basic Concepts 93

7.3 Typical Structure Models 96

7.4 ROM Layer in VLSI 97

7.5 Examples 98

7.6 Summary 107

PART Ⅲ HARDWARE IMPLEMENTATIONS 109

8 DIGITAL ANN COMPRESSOR DESIGN 109

8.1 Introduction 109

8.2 C2pL Model 110

8.3 3-2Compressor Design 111

8.4 DNN Applications 121

8.5 Summary 123

9 HYBRID PROGRAMMABLE ANN DESIGN 125

9.1 Introduction 125

9.2 Analysis and Design for PRNN 127

9.3 Improved PRNN Circuit 131

9.4 Experimental Results 135

9.5 Summary 136

Appendix A 137

INDEX 175