Behavior of Community Self-Organizing Map with Camaraderie

Abstract — In the previous study, we have proposed the Com-munity Self-Organizing Map (CSOM) that the neurons create some neuron-community according to their winning frequencies. In this study, we modify the algorithm of CSOM and propose CSOM with camaraderie. As neurons belonging to any commu-nity have the camaraderie in the community, they tend to attract each other. Therefore, the neurons belonging to any community self-organize only concentrated part of input data including a lot of noises. We apply modify CSOM and CSOM with camaraderie to clustering problem for clustering and data extraction, and investigate its behavior and its effectiveness. The efficiencies of COM with camaraderie are confirmed by several results. I