Cellular Nonlinear Networks: State of the Art and Applications - Systematic Design for Emergence in Cellular Nonlinear Networks - page 9

2.2 Typical Applications of Cellular Computers
A search done several years ago on the IEEE Xplore database reveals the following
distribution of applicative areas for cellular automata and cellular neural network
architectures (Fig. 2.1).
Fig. 2.1. Paper distribution on various applicative subjects; Based on IEEE publications
between 1988 and 2003. The number within each category is the number of papers found to
deal specifically with a certain item (e.g. VLSI) while having the words “cellular automata”
or “cellular neural network” also in the article title
The above figure reveals that most of the applications of cellular computers are
related to VLSI implementations. They are either digital implementations (custom
or reconfigurable) or mixed-signal. A notable example from the mixed-signal
category is the CNN-UM (CNN universal machine) mentioned before.
The massive parallelism of the cellular computing architecture is the more ap-
pealing feature for a VLSI implementation. The result is a fast signal-processing
engine, outperforming a conventional signal processor several orders of magni-
tudes. This is particularly effective for multi-dimensional signal processing, i.e.
image processing.
Another popular application of cellular computers is that of pattern recognition .
Several papers proposed so far the use of cellular computers as classifiers, work-
ing on a different principle than classic feed-forward neural networks. Here the
classes are associated with a finite number of attractors and the initial state
with the pattern to be recognized. Convergence towards an attractor or another
indicates the membership of the initial state pattern to a certain category. While