Universal learning networks with adaptive node functions
Kenichi Gotō, Kotaro Hirasawa, Jinglu Hu · Society of Instrument and Control Engineers of Japan · 2003
In this paper, a new architecture of neural networks is proposed named universal learning networks with adaptive node functions (ULNs with ANF), which aims at improving the learning performance and generalization ability of neural networks in terms of constructing node functions especially fitted for a certain problem. The simulation results show that the proposed method is effective for constructing networks.