Information Processing by Chemical Reaction-Diffusion Media: From Computing to Vision

Abstract

Chemical reaction-diffusion media represent information processing means fundamentally different from contemporary digital computers. Distributed character and complex nonlinear dynamics of chemical reactions inherent in the medium are the basis for large-scale parallelism and complex logical operations performed by the medium as primitives and equivalent to hundreds of binary fixed-point operations. Photo-sensitive catalysts controlling dynamics (modes of functioning) of the medium enable to easily perform input of initial data and output of computational results. It was found during the last decades that chemical reaction-diffusion media can be effectively used for solving artificial intelligence problems, such as image processing, finding the shortest paths in a labyrinth, and some other important problems that are at the same time problems of high computational complexity. Spatially non uniform control of the medium by physical stimuli and fabrication of multi level reaction-diffusion systems seem to be the promising way of enabling low cost and effective information processing devices that meet the commercial needs. Biological roots and specific neural net architecture of reaction-diffusion media seem to enable simulating some phenomena inherent in the cerebral cortex, such as optical illusions.