Cooperative information control for self-organizing maps

Abstract

This paper proposes a novel information-theoretic approach to self-organizing called cooperative information control . The new method aims to mediate between competition and cooperation among neurons by controlling information content in the neurons. Competition is realized by maximizing information content in neurons. In the process of information maximization, only a small number of neurons win the competition, while all the others are inactive. Cooperation is implemented by having neurons behave similarly to their neighbors. These two processes are unified and controlled in the framework of cooperative information control. We applied the new method to four problems: political, medical, linguistic data analysis and applied linguistic data analysis. In all the analyses, experimental results confirmed that competition and cooperation are flexibly controlled and that the method can yield a number of different neuron firing patterns to detect macro as well as micro features in input patterns.