Symbol Acquisition from Motion Patterns and Their Sparce Coding by an Associative Memory Model with Self-organizing Nonmonotonicity

Abstract

As has been claimed by Barlow [2], and reported by some recent neuro-physiological researches, at higher levels in the hierarchy of representations in the brain, sparse coding is adopted. Sparse coding is a kind of neural representation in which a very small number of neurons fire selectively. Because of the small overlaps between the firing patterns, the codes have the property of uniform metric, which may correspond to abstract symbolic representation of physical patterns [6]. We propose here a system that generates sparse codes of concepts of motions from accumulated feature vectors of observed motion patterns, by extending our previous research [11]. We propose an associative memory dynamics model with a self-organizing nonmonotonic activation function, which automatically finds out the hierarchical cluster structures in the stored motion patterns. Based on analysis of the dynamics of this model, we design an output function for the attractors, which can generate the sparse codes of the symbols of motion patterns.