SLNN: A Neural Network for Fuzzy Neural Network´s Structure Learning

A novel structure learning algorithm for fuzzy neural networks (SLNN) is presented in this paper. The neurons of SLNN are created and adapted as online learning proceeds. The learning rule of SLNN is based on Hebbian learning and a kernel winner-take-all algorithm - KWTA. KWTA not only can let SLNN be able to learn from new data but also can prevent losing the knowledge which has been learned earlier. To obtain a concise fuzzy rule, a pruning algorithm is adopted in SLNN which doesn´t disobey the basic design philosophy of fuzzy system. Simulations are performed on the primary benchmark: circle-in-the-square. Comparison with ARTMAP and BP neural network indicates that better performance is achieved