A New Dynamical Memory System Based on Chaotic Neural Networks
Kazuhiro Kojima, Koji Ito · 1998
This paper proposes a new dynamical memory system based on chaotic neural networks, and its learning method. It is demonstrated that, when no embedded pattern, i.e., unknown pattern, is applied to the system, the output pattern travels around the embedded patterns, and the traveling phases depend on a particular parameter of the networks. KEYWORDS: Associative Memory, Chaotic Dynamics, Lorenz system, Noise, Theta Rhythm. 1. INTRODUCTION In the conventional associative memory based on Liapunov stability theory[1], if an additional new pattern is applied to the network, the system must converge to 1)one of the embedded patterns, 2)the input pattern, or 3)a mixed pattern of them. On the other hand, e.g.,olfactory bulb of rabbits responds chaotically, when a novel smell is applied to it[2].This chaotic state is known as "I don't know" state. Thus, the brain has spatio-temporal dynamical memory systems. This paper proposes a new dynamical memory system based on chaotic neural networks, a...