Co-evolutionary algorithm for RBF by self-organizing population of neurons

We study the macroscopic dynamics of a three-state mean-field neural network, by using information theory principles. The results are expressed in terms of the relevant order parameters of the system, The network is composed by bilinear and biquadratic synaptic interactions, which yields an improvement of the storage and information properties of the network, measured by the mutual information and the basins of attraction. The introduction of a self-adaptive mechanism in the synapses is analyzed. The coexistence of the usual retrieval phase with a quadrupolar phase, as well as with a new informative anti-quadrupolar phase, is observed.