Abstract
A Neural Network is a nonlinear mapping system derived in reference to the brain. It is conposed of neurons and their connections, and the structure of network has close relation to the mapping capability. However, the structure has been decided by experience or by "trial and error". In this study, a decision technique for network structure was proposed. The present technique referenced just local information of the whole network and it worked in the local part of the network. Two-Spiral Problem was used to verify the learning capability of the self-organized network presented in this paper.
