Cellular and Natural Computing Models and Software Simulation - Systematic Design for Emergence in Cellular Nonlinear Networks - page 17

Reflective boundary conditions : The border cells are mirrored, i.e. the consequence

is having symmetric border properties.

In the remainder of this chapter exclusively square lattices and periodic bound-
ary conditions are used.

3.2 Major Cellular Systems Paradigms

3.2.1 The Cellular Neural Network (CNN) Model

The cellular neural network (CNN) model was proposed by Chua [27] as a practi-
cal circuit alternative to opfield and ot her type of recurrent networks. The CNN
cell is a continuous time and continuous state dynamical system with some satu-

rated nonlinearity (see (3.1)) which is well suited for implementation using analog

circuits. nlike Cs which are mostly used to prove various theories or to model
physical processes the CNN was intended from the beginning to be also a useful
signal processing paradigm.
n important step towards making this paradigm an application oriented one
was the introduction in 1

3 of the concept of CNN universal machine (CNN-
) [2]. ithin the framework of the CNN-, a CNN kernel is employed to
perform different parallel information processing tasks. Each task is associated
with a gene (i.e. a set of parameters of the synaptic interconnections) which may
be selected from a continuously growing library of more than 200 different primi-
tive genes (and tasks). Therefore one may combine various such primitives, which
are stored in an analog memory much like the instruction-code memory of digital
microprocessor. arious high-level comp utations emerge, making the CNN-M
implementation a suitable medium for computing at TerraOps computing speed.
ecent electronic implementations of the CNN-M are in fact sensor computers
[2], having the capability to sense and to process an image on the same chip.
Several generations of microelectronic chips were reported so far [31] , as well
as development tools, allowing users to program the CNN as visual microprocessor.
There is a wide range of applications, mostly in the area of image processing.
Such applications include image segmentation, image compression, fast halftoning,
contour tracking, image fusion, pattern recognition, to name just a few.
lthough initially the equilibrium dynamics of CNNs was mostly exploited for
applications, recently the nonequilibrium dynamics is employed for certain inter-
esting applications in what is currently called “computing with waves” [3,].
There is an increasing research interest around the world for the field of cellular
neural networks, most of which is reported in the roceedings of the IEEE CNNA
workshops (cellular neural networks and their applications), ISCS or IJCNN