Information Technology Reference
In-Depth Information
Two different versions of the dynamics of the simulated neural network were
investigated.
The first one involved filling the first reactor only with a solution of sodium
chlorite, while all others remained in the states with high concentration of iodine
and were colored blue (asymmetric boundary conditions). As a result of the
subsequent evolution of the system, after the links between the reactors were
activated, one observed the propagation of the wave switching from a state with a
high content of iodine into the state with its low content (i.e., color changes from
blue to colorless).
In the second version the boundary conditions were symmetrical (i.e., the first
and the last reactor were initially filled with a solution of sodium chlorite). In this
case one could observe sequential appearance of various stationary structures,
depending on the flow speed in the system of reactors (Fig.
5.31
).
In the early 1990s of the past century the group of the well-known physical
chemist John Ross at the Stanford University (USA) made an important step in
understanding the importance of related reaction-diffusion systems for information
processing. They used the network architecture of these systems in order to consider
theoretically the possibility of creating on their basis of information-logical devices
and, in particular, the implementation of Turing machines and neural networks of
the Hopfield type.
Later it was shown experimentally that neural networks based on chemical
systems have the ability to recognize simple images (see details in reference 5).
They used a bi-stable environment based on the reaction:
Fig. 5.31 Stationary structures appearing in a system of connected reaction-diffusion reactors
