Agent swarm regression network ASRN

A multi-agent system (MAS), with independent software agents interacting with each other to achieve common goals will complete concurrent distributed tasks under autonomous control. In this paper, novel RBF regression network ''agent swarm regression network ASRN" is proposed and trained by a MAS. Each neuron of the ASRN is considered as an agent, which consists of per-defined simple agent behavior set. After a sufficient number of iterations, the weights of neurons can be determined. Two sets of experiment are examined to observe the effectiveness of the proposed method.