The generation of oscillations in a biological conductance-based single-neuron model and three isolated nctwork-based models is studied using enmputer simulation. The synchronization of oscillations in a linear ch,~_in of locally coupled neural oscillators is investigated. Simulation results for linear chains employing four oscillator models are presented and their speed of synchronization is compared. A twodimensional oscillator network that attains feature-linking and pattern segmentation through temporal correlations among neural signals is presented. The network is simulated numerically for three oscillator models and the results are compared.
1. Introduction
Cyclical activities are basic characteristics of all living organisms. Neurobiologists have identified two basic types of mechanisms: single neuron that exhibits oscillations in its pulse train, which biologists refer to as "pacemakers"[ 1], and an interconnected network of neurons, where oscillations are obtained by virtue of interconnections and/or cell properties. A single neuron often possesses membrane properties that are responsible for the generation of oscillations. When coupled with other neurons, oscillations with varying properties depending on the type of intereonnecfion can be generated. Recently, synchronization of oscillations in a neural network has attracted considerable attention. It has been proposed that synchronization can solve the following tasks: 1. Phase-locking in response to global coherence in the stimuli. 2, Segmentation of incoherent stimuli in low level vision via desynchronization. 3, Feature binding i.e. connecting different attributes of the same object which appear in the mixed input.