Information Technology Reference
In-Depth Information
Actual solutions to these problems are often long and complicated descriptions
resulting in messy hardware architectures or software modules and requiring many
resources. On the other hand, the actual social demand is to develop a wide range
of intelligent but portable products (see recent tendencies described as ubiquitous
computing [12], ambient intelligence [13], organic computing [14], etc.), inte-
grated with various sensors that are available at decreasing cost prices. Since the
most important critical issue for these systems is power consumption, messy algo-
rithms and architectures are certainly not a solution. Besides, such devices are
supposed to behave naturally, i.e. perform such functional tasks which are similar
in nature to functions listed above as major challenges.
A better solution for designing such systems is to take the approach to be
developed in this topic, i.e. the
design for emergence
approach (first expressed in
[15]). Novel technologies such as nanotechnology or molecular electronics are
also better physical supports for the ideas to be developed further within this topic.
In the next we will describe what is the definition of emergence and how design
for emergence can be achieved. Then the simplest model of a natural computing
system to perform emergent computation is introduced in the form of cellular
nonlinear network (CNN). While being a generalization of the widely known cel-
lular automata (CA) model, we will see that a simple altering of this model is pos-
sible, making it compatible with the more realistic “small world” model [1].
Regarded as a potential model of universe by Zuse [16], the cellular computing
model was recently proposed by Wolfram [17] as the main ingredient of a “new
kind of science”. The praxis of this science may also benefit from the “design for
emergence” methods [15] to be further introduced.
The philosophy of designing for emergence can simply be stated as follows:
Instead of giving a description for a function to be performed, one rather
defines and determines the preconditions for emergence, i.e. a set of condi-
tions to select among many possible genes associated with basic cells.
Then, one simply picks genes within this reduced pool and observes (visu-
ally or automatically) details of the emergent behaviors and further selects
only those genes that are best fit to the application in mind.
This topic passes throughout all steps of a design for emergence process and
provides in the end (Chap. 8) several application examples where these principles
were successfully applied. In order to understand the relevance and the potential
of cellular computing models (also called cellular nonlinear networks or CNNs,
cellular arrays, etc.) the next chapter will provide and overview of the field and its
state of the art. Then, Chap. 3 introduces several of the most widely used cellular
models. Matlab simulators are provided for each of them. Moreover, in order
to cover in a unified framework a large spectrum of cellular computing arrays
(CA), a convenient taxonomy of cells is introduced in the same chapter. This tax-
onomy relies on a piecewise-linear representation of cells as nonlinear functions
rather than tables. It also allows to define
families of cells
as a narrowed search
pool for detecting emergence. So far emergence was subjectively defined, but in
order to design for emergence one needs to quantify it. A first set of measures for
Search WWH ::
Custom Search