On the Construction of Artificial Brains (eBook)

Contents 5
Prologue 9
Main Results 10
Prehistory of Out Project 14
Acknowledgement 15
The Difficulty of Modelling Artificial Barains 18
McCullogh-Pitts Model 19
Learning Nets 19
Spiking Neurons 21
Architecture of Vision 22
The Steps of the Construction Process 24
Summary 25
Information Processing in Nets with Constant Synapses 27
Generic Signal Equations for Pulse Neurons and Synapses 28
Partitions and Their Time Development 29
Experiments with Constant Synapses 31
Entropy and Transfer Function of a Net 37
Operating Range of a Net 39
Pulse Rates 39
Resoluation and Net Size 42
Application Potential 46
Limited Simulation Time 50
Summary 50
Theory of Nets with Constant or Dynamic Synapses 52
Derivation of the Signal Energy 52
Temporal Mean and Spatial Mean 56
Determination of the Frequency Distribution 58
Summary 63
Macro-Dynamics of Nets with Constant Synapses 64
Known Synapses 64
Known Distribution of Synapses 69
Agreement of Theory with Experiment 71
Lock of Correlation 84
Determining the Signal Energy and Entropy by Pulse Rates 84
Summary 86
Information Processing with Dynamic Synapses 87
The Types of Solutions of Synaptic Equations 88
Synchronisation of Neurons 89
Segmentation per Synchronisation 95
Calcylation of Pulse Differences and Sums 97
Simple Applications 100
Time Coding and Correlation 104
Entropy and State Space 105
Preliminary Considerations on the Statistics of Synchronisation 107
Summary 107
Nets for Feature Detection 109
Overview of Visual System 111
Simple Cells 112
Creation of Detector Profiles for Gabor Wavelets 115
Experimental Check 119
Summary 121
Nets for Feature Recognition 123
Principles of Object Recognition 125
Net Architecture for Robust Feature Recognition 127
Feature Recogniser 130
Selectivity 132
Orthogonality of Rotation 135
Invariance of Function as to Brightness 136
Invariance of Function as to Form and Mimic 136
Generating Object Components through Binding of Features 140
Summary 142
Nets for Robust Head Detection 144
Results of Head Detection 145
Next Steps 149
Summary 151
Extensions of the Vision Architecture 152
Distance-Invariant Feature Pyramid 152
The Inner Screen 164
Summary 173
Look-out 175
Data Format of the Brain 175
Self-Organisation 176
Learning 178
Invariant Object Recognition 179
Structured Memory Domains 181
Summary 183
Preliminary Considerations on the Microelectronic Implementation 185
Equivalent Representations 185
Microelectronic Implementations 186
Models of Neurons and Synapses 188
Elementary Circuits for Neurons, Synapses, and Photosensors 197
Neuron 197
Adaptive Synapses 207
Photosensors 212
DA-Converters and Analogue Image Storage 232
Summary 233
Simulation of Microelectronic Neural Circuits and Systems 234
Modelling of Neurons and Synapses 235
Results of Modelling 236
Notes on the Simulation Procedure 238
Summary 247
Architecture and Chip Design of the Feature Recognizer 248
Chip Architecture of the Feature Recogniser 249
Interfaces for Test and Readout 250
Chip Design and layout 253
Demonstrator and Measurement Results 253
Summary 255
Architecture and Chip Design of the Feature Detector 256
Digital Representation of the Feature Detection 256
VLSI Design of a Neural Processor and Router Circuit 258
Demonstration of the Feature Detection 264
Sumary 268
3D Stacking Technology 269
Fundamentals of 3D Stacking 271
Processing Steps for 3D Stacking 273
Assembly 292
Electrical Properties of 3D Interconnections 296
Summary 298
Architecture of First Generation Vision Cube 299
Imager Chip 299
AER Chip 300
Size of the NPU Array and Memory Requirements 304
Chip for Feature Detection 310
Chip for the Feature Recogniser 312
Summary 321
End 324
Appendix 326
Simulator for Chapters 3,4 326
Axon model by Hodgkin and Huxley 333
Basic Transistor Circuits 337
Optical Generation 343
References 346
Index 353