Informational processing on time sequential input with coupled nonlinear oscillators

Abstract

In living organisms, informational processing is carried out in highly manner. Time-dependence in the response of sensory organs is considered to play an essential role to take in the rich information from the outer environment. After the transduction of the outer information, the time-dependent, rich information is transformed into valuable information with neural net. The individual element of the neural network, i.e., the living neuronal cell, is regarded as a kind of excitable device, and works along the time axis. In the present paper, we propose a novel idea to construct dynamic artificial neural-net, mimicking the essentials of biological neural-net. We consider the network of nonlinear oscillators with linear coupling. In the network of N-coupled nonlinear oscillators, entrained modes with (N-1)! different states exist, even under the framework only to consider N-phase solution. In addition, by use of coupled oscillators, we can realize the time dependent function, such as remembering and forgetting. We will also show that the memory capacity of memory in the N-coupled oscillators can be much more than (N-1)!, by considering various modes of entrainment among the oscillators.