Analyses of dynamical coherence in the pattern recognition model of oscillator neural networks, Holovision

As a neural mechanism of recognition in the brain, generation of coherence through synchronization has drawn attention experimentally and theoretically. The authors have proposed the principle of generation of dynamical coherence by synchronization and based on this principle, the pattern recognition model composed of two centers of oscillator neural networks has been proposed. Flexible and stable ability of pattern recognition is obtained in the model, which is called Holovision. In the present paper, the analyses of bifurcation of coherent states as synchronization is studied on the authors\´ recent model. "In-phase approximation" first clarifies the dynamical states of individual units as a collective mode. Next, the bifurcation diagram obtained by computer experiments shows the significance of reciprocal information flow between two kinds of information, a percept and symbols. It is concluded that generation of coherence among local heterogeneous activities over the two centers critically unifies the information in attainment of pattern recognition.