Symbolic Memory of Motion Patterns by an Associative Memory Dynamics with Self-organizing Nonmonotonicity

We previously proposed a memory system of motion patterns[4] using an assotiative memory model. It forms symbolic representations of motion patterns based on correlations by utilizing bifurcations of attractors depending on the parameter of activation nonmonotonicity. But the parameter had to be chosen appropreately to some degree by manual. We propose here a way to provide the paremeter with selforganizing dynamics along with the retrieval of the associative momory. Attractors of the parameter are discrete states representing the hierarchical correlations of the stored motion patterns.