Influence of parameter deviations in an associative chaotic neural network

Influence of parameter deviations in an associative chaotic neural network is analyzed in this paper. The network is composed of chaotic model neurons whose parameter values have small deviations among the neurons in the network. The associative chaotic neural network with 16 chaotic neurons is analyzed in this paper. Three orthogonal pattern vectors are stored by setting synaptic weights of the network with conventional auto-associative matrix. Comparisons of the network with parameter deviations and the network with common parameters are made in the retrieval characteristics and in the dynamical property. In order to make such comparison, the nominal values of the parameters of the network are set so that the networks shows periodic retrieval of one of the stored patterns when the parameter values are common to every neuron in the network. The network with the parameter deviations among the neurons shows aperiodic associative dynamics even if the nominal values of the parameters are set to show periodic behavior. The aperiodic associative dynamics of the network with the parameter deviations can be characterized by the index for the orbital instability.