Agent swarm classification network ASCN

In this paper we introduced a new RBF classification network - "agent swarm classification network ASCN", which is trained by a multi-agent system (MAS) approach. MAS can be regarded as a swarm of independent software agents interact with each other to achieve common goals, complete concurrent distributed tasks under autonomous control. By treating each neuron as an agent, the weights of neurons can be determined through a set of pre-defined simple agent behavior. Three sets of experiments are examined to observe the effectiveness of the proposed method.