Fuzzy neural fusion techniques for industrial applications

Abstract

Three different fuzzy neural fusion methods for real world problems are presented with results for comparison, The fuzzy controlled back propagation converges faster in comparison to the conventional back propagation. The cascade system fuzzy step net extends the ability of integrating neural learning with fuzzy logic statements, giving an insight view over the complexity of separation of output classes. The neural network based fuzzy system generator FuNe I analyses the training data automatically indicating superfluous input and generating a knowledge base. All three systems are tested with Iris flower species classification data and applied to industrial problems.