A Self Adaptive Incremental Learning Fuzzy Neural Network Based on the Influence of a Fuzzy Rule

In a fuzzy neural network, a fuzzy rule may be active in early stage, then the contribution of the rule to system become small. In this paper, A Self Adaptive incremental learning Fuzzy Neural Network Based on the Influence of a Fuzzy Rule (SAIL-FNN) is developed. In SAFIS, the concept of "influence" of a fuzzy rule is introduced and fuzzy rules are added or removed based on the influence for the input data received so far. Furthermore, the "Significance" of a neuron is linked to the learning accuracy. Only the value of significance of a rule is larger than a threshold, and then one rule may consider to be added. Else the rule is updated using an extended kalman filter (EKF) scheme. An experiment validates our theoretical results. The results indicate that the SAIL-FNN algorithm can provide comparable generalization performance with a considerably reduced network size and training.