Cellular Nonlinear/Neural Network (CNN) technology is both a revolutionary concept and an experimentally proven new computing paradigm. Analogic cellular computers based on CNNs are set to change the way analog signals are processed. This unique undergraduate level textbook includes many examples and exercises, including CNN simulator and development software accessible via the Internet. It is an ideal introduction to CNNs and analogic cellular computing for students, researchers and engineers from a wide range of disciplines. Leon Chua, co-inventor of the CNN, and Tamàs Roska are both highly respected pioneers in the field.
Author(s): Leon O. Chua, Tamas Roska
Edition: 1st
Publisher: Cambridge University Press
Year: 2002
Language: English
Pages: 410
Cover......Page 1
Half-title......Page 3
Title......Page 5
Copyright......Page 6
Dedication......Page 7
Contents......Page 9
Acknowledgements......Page 13
Scenario......Page 15
The textbook......Page 17
New developments......Page 19
2.1 Basic notation and definitions......Page 21
2.2.1 Vector and matrix representation and boundary conditions......Page 28
Boundary conditions......Page 29
Vector differential equation......Page 31
2.2.2 Existence and uniqueness of solutions......Page 32
2.2.3 Boundedness of solutions......Page 37
Cloning template representation......Page 39
2.2.5 Three simple CNN classes......Page 41
2.2.6 Synaptic signal flow graph representation......Page 44
EDGE: binary edge detection template......Page 49
3.1.2 The EDGEGRAY CNN......Page 54
EDGEGRAY: gray-scale edge detection template......Page 55
3.2 Three quick steps for sketching the shifted DP plot......Page 63
3.3 Some other useful templates......Page 64
3.3.1 CORNER: convex corner detection template......Page 65
3.3.2 THRESHOLD: gray-scale to binary threshold template......Page 68
FILBLACK: Gray-scale to black CNN......Page 72
FILWHITE: Gray-scale to white CNN......Page 73
3.3.4 LOGNOT: Logic NOT and set complementation…......Page 74
3.3.5 LOGOR: Logic OR and set union (disjunction) template......Page 75
3.3.6 LOGAND: Logic AND and set intersection (conjunction) template......Page 78
3.3.7 LOGDIF: Logic difference and relative set complement (P \ P = P – P) template......Page 82
3.3.8 SHIFT: Translation (by 1 pixel-unit) template......Page 86
3.3.9 CONTOUR-1: Contour detection template......Page 93
3.3.10 EROSION: Peel-if-it-doesn’t-fit Template......Page 98
3.3.11 DILATION: Grow-until-it-fits template......Page 105
4.1 Integration of the standard CNN differential equation......Page 114
4.2 Image input......Page 115
4.3 Software simulation......Page 116
4.4 Digital hardware accelerators......Page 124
4.5 Analog CNN implementations......Page 125
4.6 Scaling the signals......Page 127
4.7 Discrete-time CNN (DTCNN)......Page 128
5.1 Binary and universal CNN truth table......Page 129
5.2 Boolean and compressed local rules......Page 136
Computer-aided method for proving local rules......Page 137
5.3 Optimizing the truth table......Page 138
6.1 The complete stability phenomenon......Page 153
6.2 Explicit CNN output formula......Page 154
6.3 Proof of completely stable CNN theorem......Page 156
6.4 The primary CNN mosaic......Page 169
6.5 Explicit formula for transient waveform and settling time......Page 170
6.6 Which local Boolean functions are realizable by uncoupled CNNs?......Page 175
6.7 Geometrical interpretations......Page 176
6.8 How to design uncoupled CNNs with prescribed Boolean functions......Page 180
6.9 How to realize non-separable local Boolean functions?......Page 187
7 Introduction to the CNN Universal Machine......Page 197
7.2 Set inclusion......Page 198
7.3 Translation of sets and binary images......Page 202
7.4 Opening and closing and implementing anymorphological operator......Page 204
7.5 Implementing any prescribed Boolean transition function by not more than 256 templates......Page 209
7.6 Minimizing the number of templates when implementing any possible Boolean transition function......Page 212
7.7 Analog-to-digital array converter......Page 215
8.2 An oscillatory CNN with only two cells......Page 219
8.3 A chaotic CNN with only two cells and one sinusoidal input......Page 224
8.4 Symmetric A template implies complete stability......Page 228
8.5 Positive and sign-symmetric A template implies complete stability......Page 233
8.6 Positive and cell-linking A template implies complete stability......Page 238
8.7 Stability of some sign-antisymmetric CNNs......Page 245
A Appendix to Chapter 8......Page 250
LaSalle’s invariance principle......Page 252
9 The CNN Universal Machine (CNN-UM)......Page 253
9.1.1 The extended standard CNN universal cell......Page 254
Why stored programmability is possible?......Page 256
9.2 A simple example in more detail......Page 258
9.3 A very simple example on the circuit level......Page 260
The flow diagram of the algorithm and the templates......Page 261
The macro code of the algorithm......Page 262
The functional circuit level schematics of an extended cell......Page 263
The content of the global analogic programming unit (GAPU)......Page 264
9.4 Language, compiler, operating system......Page 268
10.1 Various design techniques......Page 272
10.2 Binary representation, linear separability, and simple decomposition......Page 274
10.3 Template optimization......Page 278
10.4 Template decomposition techniques......Page 279
11.1 Linear image processing with B templates is equivalent to spatial convolution with FIR kernels......Page 281
11.2 Spatial frequency characterization......Page 283
11.4 Linear image processing with A and B templates is equivalent to spatial convolution with IIR kernels......Page 286
12 Coupled CNN with linear synaptic weights......Page 290
12.1 Active and inactive cells, dynamic local rules......Page 292
12.2 Binary activation pattern and template format......Page 297
12.3.2 Local rules and binary activation pattern......Page 298
12.3.4 System of inequalities and optimal solution......Page 299
12.4 The connectivity problem......Page 300
12.4.2 Local rules and binary activation pattern......Page 301
12.4.3 Template type and template form......Page 302
12.4.4 System of inequalities and optimal solution......Page 303
13 Uncoupled standard CNNs with nonlinear synaptic weights......Page 304
13.1 Dynamic equations and DP plot......Page 305
Gray-scale contour detector......Page 306
14.1 Dynamic equations......Page 310
14.2 Motion analysis – discrete time and continuous time image acquisition......Page 311
Generating the difference picture in continuous time mode......Page 313
15 Visual microprocessors – analog and digital VLSI implementation of the CNN Universal Machine......Page 317
15.1 The analog CNN core......Page 318
15.2 Analogic CNN-UM cell......Page 324
15.3 Emulated digital implementation......Page 326
15.4 The visual microprocessor and its computational infrastructure......Page 327
15.5 Computing power comparison......Page 332
16 CNN models in the visual pathwayand the ‘‘Bionic Eye”......Page 334
16.1 Receptive field organization, synaptic weights, and cloning template......Page 335
16.2 Some prototype elementary functions and CNN models of the visual pathway......Page 336
The triad synapse action......Page 338
Directional selectivity......Page 339
Length tuning......Page 340
A simple visual illusion......Page 342
16.3 A simple qualitative ‘‘engineering” model of a vertebrate retina......Page 343
The cell prototype......Page 344
Receptive field organization types (RF)......Page 345
Multilayer CNN for receptive field interactions......Page 346
The structure of a prototype retinal model......Page 348
16.4 The ‘‘Bionic Eye” implemented on a CNN Universal Machine......Page 352
3 Characteristics and analysis of simple CNN templates......Page 353
6 Uncoupled CNNs: unified theory and applications......Page 354
7 Introduction to the CNN universal machine......Page 355
8 Back to basics: Nonlinear dynamics and complete stability......Page 356
10 Template design tools......Page 357
14 Standard CNNs with delayed synaptic weights and motion analysis......Page 358
16 CNN models in the visual pathway and the ‘‘Bionic Eye”......Page 359
1988–1990......Page 362
1991–1992......Page 363
1993–1994......Page 364
1995–1996......Page 367
1997–1998......Page 369
1999......Page 372
Exercise 2.2 (Hexagonal neighborhood)......Page 375
Exercise 3.1 (Separate connected objects)......Page 376
Exercise 3.2 (EDGE–CORNERDETECTION comparison)......Page 377
Exercise 3.3 (Main group of points)......Page 378
Exercise 6.1 (Crossword puzzle endings)......Page 379
Exercise 8.2 (Reaction–diffusion equations)......Page 380
Exercise 8.3 (Surface interpolation)......Page 382
Exercise 8.4 (Black pixel count)......Page 383
Exercise 9.1 (Roughness measurement)......Page 384
Exercise 9.3 (Concavity orientation)......Page 386
Exercise 9.4 (Improved concavity orientation)......Page 387
Exercise 9.5 (Curvature)......Page 388
Exercise 9.6 (Absolute value)......Page 389
Exercise 9.7 (X and O segmentation)......Page 390
Exercise 9.8 (QCA simulation)......Page 391
Exercise 10.1 (Template design)......Page 392
Exercise 12.1 (Distance classification)......Page 393
Exercise 12.2 (Arc detection)......Page 394
Exercise 12.3 (Detect forks)......Page 395
Exercise 12.4 (Locate small ellipses)......Page 397
Exercise 13.1 (Linear morph)......Page 399
Exercise 13.4 (Chaotic cell)......Page 401
Appendix C: CANDY, a simulator for CNN templates and analogic CNN algorithms......Page 403
Index......Page 404
