A LEARNING-MODEL FOR OSCILLATORY NETWORKS

A learning model for coupled oscillators is proposed. The proposed lea rning rule takes a simple form by which the intrinsic frequencies of t he component oscillators and the coupling strength between them are ch anged according to the effects of the input signals on the dynamics of the oscillator. In the learning mode, each component oscillator recei ves a teacher signal of desired phase and frequency, and a desired par ameter set for generating the desired pattern is acquired by the propo sed learning rule. It is known that the basic locomotor patterns of ma ny living bodies are generated by coupled neural oscillators. The prop osed learning rule could be a learning model used by such neural syste ms to acquire an adequate parameter set for generating a desired locom otor pattern. (C) 1998 Elsevier Science Ltd. All rights reserved.