Neural Networks and Micromechanics (eBook)

FM 1
1 10
Chapter 1: Introduction 10
References 13
2 15
Chapter 2: Classical Neural Networks 15
Neural Network History 15
McCulloch and Pitts Neural Networks 15
Hebb Theory 16
Perceptrons 18
Neural Networks of the 1980s 20
Modern Applications of Neural Network Paradigms 23
Hopfield Neural Networks 23
Adaptive Resonance Theory (ART) 24
Self-Organizing Feature Map (SOFM) Neural Networks 25
Cognitron and Neocognitron 26
Backpropagation 27
Adaptive Critic Design 28
RTC, RSC, LIRA, and PCNC Neural Classifiers 29
References 29
3 34
3: Neural Classifiers 34
RTC and RSC Neural Classifiers for Texture Recognition 34
Random Threshold Neural Classifier 36
Random Subspace Classifier 38
Encoder of Features 39
LIRA Neural Classifier for Handwritten Digit Recognition 40
Rosenblatt Perceptrons 41
Description of the Rosenblatt Perceptron Modifications 42
Mask Design 44
Image coding 45
Training procedure 45
Recognition procedure 47
LIRA-Grayscale Neural Classifier 48
Handwritten Digit Recognition Results for Lira-binary 49
Handwritten Digit Recognition Results for LIRA-Grayscale 50
Discussion 51
References 52
4 54
Chapter 4: Permutation Coding Technique for Image Recognition System 54
Special- and General-Purpose Image Recognition Systems 54
Random Local Descriptors 56
General Purpose Image Recognition System Description 57
Computer Simulation 61
Permutation Coding Neural Classifier (PCNC) 61
PCNC structure 61
Feature extractor 62
Encoder 63
PCNC Neural Classifier Training 71
Results Obtained on the MNIST Database 72
Results Obtained on the ORL Database 73
References 78
5 81
Chapter 5: Associative-Projective Neural Networks (APNNs) 81
General Description of the Architecture 81
Neuron, the Training Algorithms 81
Neural Fields 84
Input Coding and the Formation of the Input Ensembles 89
Local Connected Coding 89
Coding the numbers and sets of the numerical parameters 91
Code normalization 92
Shift Coding 94
Centering the shift code 98
Application of shift coding 102
Functions of Neural Ensembles 102
Methods of Economical Presentation of the Matrix of Synaptic Weights (Modular Structure) 104
Stochastic not fully connected networks 104
Constructing modular neural networks 107
Conclusion 108
References 109
6 111
Chapter 6: Recognition of Textures, Object Shapes, and Handwritten Words 111
Recognition of Textures 111
Extraction of Texture Features 111
The Coding of Texture Features 112
Texture Recognition 113
The Experimental Investigation of the Texture Recognition System 115
Texture Recognition with the Method of Potential Functions 117
Recognition of Object Shapes 119
Features for Complex Shape Recognition 120
Experiments with Algorithms of Complex Shape Recognition 122
Recognition of Handwritten Symbols and Words 123
Features Utilized for the Recognition of Handwritten Words 123
The algorithm of line thinning 124
Algorithm of line thickening 125
Extraction of informative features 126
Information coding 126
Coding of binary features 127
Coding the feature position on the image 127
The experimental results 129
Optimization of the feature set 129
Conclusion 133
References 134
7 136
Chapter 7: Hardware for Neural Networks 136
Neurocomputer NIC 136
Description of the Block Diagram of the Neurocomputer 136
The Realization of the Algorithm of the Neural Network on the Neurocomputer 138
Neurocomputer B-512 139
The Designation of the Neurocomputer Emulator 140
The Structure of the Emulator 140
The Block Diagram of B-512 143
Conclusion 145
References 145
8 146
Chapter 8: Micromechanics 146
The Main Problems of Microfactory Creation 146
General Rules for Scaling Down Micromechanical Device Parameters 150
The Analysis of Micromachine Tool Errors 153
Thermal Expansion 153
Rigidity 155
Compression (or extension) of the bar 155
Bending of the bar. Case 1 156
Bending of the bar. Case 2 157
Torsion of the bar 157
Forces of Inertia 158
Force of inertia. Linear movement with uniform acceleration 158
Centrifugal force 159
Magnetic Forces 160
Electrostatic Forces 160
Viscosity and Velocity of Flow 161
Pneumatic and hydraulic forces 161
Forces of surface tension (capillary forces) 163
Mass Forces 164
Forces of Cutting 164
Elastic Deformations 166
Vibrations 167
The First Prototype of the Micromachine Tool 169
The Second Prototype 172
The Second Micromachine Tool Prototype Characterization 174
Positional Characteristics 175
Geometric Inspection 182
Errors that Do Not Decrease Automatically 183
Methods of Error Correction 183
The method of ``lever.´´ 184
Micromachining center based on parallograms 185
Parallel micromanipulators 185
Adaptive Algorithms 186
Adaptive Algorithms Based on a Contact Sensor 186
Possible Applications of Micromachine Tools 190
The Problem of Liquid and Gas Fine Filtration 190
Design of Filters with a High Relation of Throughput to Pressure Drop 191
An Example of Filter Design 193
The Problems of Fine Filter Manufacturing 193
The Filter Prototype Manufactured by the Second Micromachine Tool Prototype 194
Case Study 194
Conclusion 196
References 197
9 200
Chapter 9: Applications of Neural Networks in Micromechanics 200
Neural-Network-Based Vision System for Microworkpiece Manufacturing 200
The Problems of Adaptive Cutting Processes 201
Permutation Coding Neural Classifier 202
Feature Extractor 202
Encoder 204
Neural Classifier 206
Results 207
References 208
10 209
Chapter 10: Texture Recognition in Micromechanics 209
Metal Surface Texture Recognition 209
Feature Extraction 211
Encoder of Features 211
Results of Texture Recognition 212
References 213
11 214
Chapter 11: Adaptive Algorithms Based on Technical Vision 214
Microassembly Task 214
LIRA Neural Classifier for Pin-Hole Position Detection 219
Neural Interpolator for Pin-Hole Position Detection 220
Discussion 223
References 224