Dynamical memories based on inter-module Hebbian correspondences with the chaotic neural network modules
A. Sano · 2003
Each memory process of the neural network is not always divisible. These memories represent by the interact with one another, supposing it is represented by nonlinear dynamics. In this article, the interacting memory process is studied in our two-module chaotic neural network model with the Hebbian learning. Internal representation of the chaotic model is classified as two types of dynamics in ordered periodic "I know" state or high-dimensional chaotic "I don't know" state. It is found that the novel periodic "I know" state is autonomously generated in the Hebbian learning process. Moreover, the inter-module coupling against the learned Hebbian correspondences also gives a novel "I know" state. These results suggest the existence of novel memories or functions generated by the interaction in the neural networks or the brain.