Ramacher / Malsburg On the Construction of Artificial Brains

1. Auflage 2010
ISBN: 978-3-642-00189-5
Verlag: Springer
Format: PDF
Kopierschutz: Wasserzeichen (»Systemvoraussetzungen)

E-Book, Englisch, 370 Seiten

This book presents a first generation of artificial brains, using vision as sample application. An object recognition system is built, using neurons and synapses as exclusive building elements. The system contains a feature pyramid with 8 orientations and 5 resolution levels for 1000 objects and networks for binding of features into objects. This vision system can recognize objects robustly in the presence of changes in illumination, deformation, distance and pose (as long as object components remain visible). The neuro-synaptic network owes its functional power to the introduction of rapidly modifiable dynamic synapses. These give a network greater pattern recognition capabilities than are achievable with fixed connections. The spatio-temporal correlation structure of patterns is captured by a single synaptic differential equation in a universal way. The correlation can appear as synchronous neural firing, which signals the presence of a feature in a robust way, or binds features into objects. Although in this book we can present only a first generation artificial brain and believe many more generations will have to follow to reach the full power of the human brain, we nevertheless see a new era of computation on the horizon. There were times when computers, with their precision, reliability and blinding speed, were considered to be as superior to the wet matter of our brain as a jet plane is to a sparrow. These times seem to be over, given the fact that digital systems inspired by formal logic and controlled algorithmically - today's computers - are hitting a complexity crisis. A paradigm change is in the air: from the externally organised to the self-organised computer, of which the results described in this book may give an inkling.

Autoren/Hrsg.

Ramacher, Ulrich

Malsburg, Christoph

Inhaltsverzeichnis

1;Contents;5
2;Prologue;9
2.1;Main Results;10
2.2;Prehistory of Out Project;14
2.3;Acknowledgement;15
3;The Difficulty of Modelling Artificial Barains;18
3.1;McCullogh-Pitts Model;19
3.2;Learning Nets;19
3.3;Spiking Neurons;21
3.4;Architecture of Vision;22
3.5;The Steps of the Construction Process;24
3.6;Summary;25
4;Information Processing in Nets with Constant Synapses;27
4.1;Generic Signal Equations for Pulse Neurons and Synapses;28
4.2;Partitions and Their Time Development;29
4.3;Experiments with Constant Synapses;31
4.4;Entropy and Transfer Function of a Net;37
4.5;Operating Range of a Net;39
4.6;Pulse Rates;39
4.7;Resoluation and Net Size;42
4.8;Application Potential;46
4.9;Limited Simulation Time;50
4.10;Summary;50
5;Theory of Nets with Constant or Dynamic Synapses;52
5.1;Derivation of the Signal Energy;52
5.2;Temporal Mean and Spatial Mean;56
5.3;Determination of the Frequency Distribution;58
5.4;Summary;63
6;Macro-Dynamics of Nets with Constant Synapses;64
6.1;Known Synapses;64
6.2;Known Distribution of Synapses;69
6.3;Agreement of Theory with Experiment;71
6.4;Lock of Correlation;84
6.5;Determining the Signal Energy and Entropy by Pulse Rates;84
6.6;Summary;86
7;Information Processing with Dynamic Synapses;87
7.1;The Types of Solutions of Synaptic Equations;88
7.2;Synchronisation of Neurons;89
7.3;Segmentation per Synchronisation;95
7.4;Calcylation of Pulse Differences and Sums;97
7.5;Simple Applications;100
7.6;Time Coding and Correlation;104
7.7;Entropy and State Space;105
7.8;Preliminary Considerations on the Statistics of Synchronisation;107
7.9;Summary;107
8;Nets for Feature Detection;109
8.1;Overview of Visual System;111
8.2;Simple Cells;112
8.3;Creation of Detector Profiles for Gabor Wavelets;115
8.4;Experimental Check;119
8.5;Summary;121
9;Nets for Feature Recognition;123
9.1;Principles of Object Recognition;125
9.2;Net Architecture for Robust Feature Recognition;127
9.3;Feature Recogniser;130
9.4;Selectivity;132
9.5;Orthogonality of Rotation;135
9.6;Invariance of Function as to Brightness;136
9.7;Invariance of Function as to Form and Mimic;136
9.8;Generating Object Components through Binding of Features;140
9.9;Summary;142
10;Nets for Robust Head Detection;144
10.1;Results of Head Detection;145
10.2;Next Steps;149
10.3;Summary;151
11;Extensions of the Vision Architecture;152
11.1;Distance-Invariant Feature Pyramid;152
11.2;The Inner Screen;164
11.3;Summary;173
12;Look-out;175
12.1;Data Format of the Brain;175
12.2;Self-Organisation;176
12.3;Learning;178
12.4;Invariant Object Recognition;179
12.5;Structured Memory Domains;181
12.6;Summary;183
13;Preliminary Considerations on the Microelectronic Implementation;185
13.1;Equivalent Representations;185
13.2;Microelectronic Implementations;186
13.3;Models of Neurons and Synapses;188
14;Elementary Circuits for Neurons, Synapses, and Photosensors;197
14.1;Neuron;197
14.2;Adaptive Synapses;207
14.3;Photosensors;212
14.4;DA-Converters and Analogue Image Storage;232
14.5;Summary;233
15;Simulation of Microelectronic Neural Circuits and Systems;234
15.1;Modelling of Neurons and Synapses;235
15.2;Results of Modelling;236
15.3;Notes on the Simulation Procedure;238
15.4;Summary;247
16;Architecture and Chip Design of the Feature Recognizer;248
16.1;Chip Architecture of the Feature Recogniser;249
16.2;Interfaces for Test and Readout;250
16.3;Chip Design and layout;253
16.4;Demonstrator and Measurement Results;253
16.5;Summary;255
17;Architecture and Chip Design of the Feature Detector;256
17.1;Digital Representation of the Feature Detection;256
17.2;VLSI Design of a Neural Processor and Router Circuit;258
17.3;Demonstration of the Feature Detection;264
17.4;Sumary;268
18;3D Stacking Technology;269
18.1;Fundamentals of 3D Stacking;271
18.2;Processing Steps for 3D Stacking;273
18.3;Assembly;292
18.4;Electrical Properties of 3D Interconnections;296
18.5;Summary;298
19;Architecture of First Generation Vision Cube;299
19.1;Imager Chip;299
19.2;AER Chip;300
19.3;Size of the NPU Array and Memory Requirements;304
19.4;Chip for Feature Detection;310
19.5;Chip for the Feature Recogniser;312
19.6;Summary;321
20;End;324
21;Appendix;326
21.1;Simulator for Chapters 3,4;326
21.2;Axon model by Hodgkin and Huxley;333
21.3;Basic Transistor Circuits;337
21.4;Optical Generation;343
22;References;346
23;Index;353