Mechanism for forming internal correlation when concepts are generated on a neural network with connections produced by Fourier holography

Abstract

The interaction of two competing mechanisms that participate in the formation of a concept is considered in developing an approach to the implementation of inductive generalization in a Fourier-holography setup as a method of generating abstract concepts on the basis of sensory-pattern processing. These mechanisms are nonlinear iterative mapping in the correlation plane, which breaks down the internal correlation of the concept, and diffraction on a rerecorded hologram, which restores the internal correlation. It is shown that the characteristics of the index (reference) pattern that represents reality and the characteristics of the hologram, determined by the properties of the recording medium and by the method of recording it, have an effect on the formation of the internal correlation of the generated concept that prevails over the characteristics of the iterative mapping.