Information Technology Reference
In-Depth Information
Fig. 4.10 Vibrational mode of the Belousov-Zhabotinsky medium
to the point of intersection of the isoclines. During this process, the concentrations
of the activator and of the inhibitor of the reaction change periodically. When
taking into account the diffusion, the oscillatory nature of the regime is kept, but the
evolution of the medium becomes more complex (see next chapter). A relatively
simple situation where one can observe the vibrational mode of the Belousov-
Zhabotinsky medium, into which an arbitrary image is introduced, is shown in
Fig.
4.10
. The original image in the medium consequently passes through the stages
“negative-positive-negative
” etc.
An important feature of an information medium of the Belousov-Zhabotinsky
type is its neural network architecture.
Qualitatively, the Belousov-Zhabotinsky medium can be regarded as a realiza-
tion of a neural network (Fig.
4.7
), where:
...
Each elementary volume of the medium can be thought of as a simple processor.
The dynamics of the processor and the operations it executes are defined only by
nonlinear kinetics of reactions occurring in the microvolume.
Processes occurring in microvolumes are related via short-range (diffusion)
interactions. Specifically, each microscopic volume is linked by diffusion with
any other microvolume of the environment. But, because of the low rate of
diffusion, interaction between volumes occurs with a delay and with attenuation
proportional to the distance between them.
Depending on the state of the medium (concentration of the reaction components
and temperature) and external excitation, the system described by reaction-
diffusion equations taking diffusion into account can operate in different
dynamic regimes.
It should be noted that distributed neural networks are more rigorously described
by a system of integro-differential equations that cannot be, in general, reduced to a
reaction-diffusion equations. Nevertheless, under certain rather relaxed assump-
tions, these two models are adequate.
