《神经元网络系统设计方法 英文》

1INTRODUCTION1

1.1 ANNs for Pattern Recognition1

1.2 Neural Network Model3

1.3 ANN Architecture6

1.4 Hardware Implementation of ANN7

1.5 A VLSI System Design Methodology10

PARTⅠPARALLEL ANN MODELS13

2AN UNSUPERVISED LEARNING MODEL13

2.1 Introduction13

2.2 Fuzzy Clustering Neural Networks17

2.3 Parallel FCNN Architecture20

2.4 Experimental Results21

2.5 Summary24

3 A SUPERVISED TRAINING MODEL29

3.1 Introduction29

3.2 Linear Separability Analysis31

3.3 Layer Adaptation Approach35

3.4 Experiment:Pattern Recognition39

3.5 Comparisons41

3.6 Summary42

4 A NEURAL-LIKE NETWORK MODEL45

4.1 Introduction45

4.2 FRNN Computing Model49

4.3 FRNN Architecture51

4.4 Case Studies52

4.5 Summary54

PART Ⅱ VLSI ARCHITECTURES57

5A PARALLEL ARCHITECTURE IMPLEMENTED BYSYSTOLIC ARRAYS57

5.1 Introduction57

5.2 FCNN Architecture58

5.3 Mapping Policies63

5.4 Typical SA Structures65

5.5 Mapping FCNN onto SA68

5.6 Summary70

6 A PIPELINED ARCHITECTURE BASED ON WINDOW OPERATION73

6.1 Introduction73

6.2 Pipelined Architecture:Window Operation73

6.3 Window Implementation80

6.4 Case Studies84

6.5 Performance Analysis89

6.6 Summary92

7 A SIMPLIFIED ARCHITECTURE USING A PRIORI KNOWLEDGE93

7.1 Introduction93

7.2 Basic Concepts93

7.3 Typical Structure Models96

7.4 ROM Layer in VLSI97

7.5 Examples98

7.6 Summary107

PART ⅢHARDWARE IMPLEMENTATIONS109

8 DIGITAL ANN COMPRESSOR DESIGN109

8.1 Introduction109

8.2 C2pL Model110

8.3 3-2Compressor Design111

8.4 DNN Applications121

8.5 Summary123

9 HYBRID PROGRAMMABLE ANN DESIGN125

9.1 Introduction125

9.2 Analysis and Design for PRNN127

9.3 Improved PRNN Circuit131

9.4 Experimental Results135

9.5 Summary136

Appendix A137

INDEX175