Title :
Takagi-Sugeno Fuzzy System Based Hierarchical Hybrid Fuzzy-Neural Networks
Author_Institution :
Res. Center of Fuzzy Syst., Beijing Normal Univ., Zhuhai, China
Abstract :
Takagi-Sugeno (T-S) fuzzy system was merged into Hierarchical Hybrid Fuzzy-Neural Networks (HHFNN) and homogeneous linear function of input variables was employed in the THEN part of fuzzy rules of T-S fuzzy systems. A new training algorithm for this model was also proposed. The parameters consist of the coefficients of homogeneous linear functions and the weights and bias terms of upper neural network. This proposed model has fewer parameters than standard BP network under the same conditions, and outperforms Mamdani fuzzy system based HHFNN and standard BP network in accuracy and error-descent speed according to two simulation examples.
Keywords :
backpropagation; fuzzy neural nets; fuzzy systems; BP network; Takagi-Sugeno fuzzy system; hierarchical hybrid fuzzy-neural networks; homogeneous linear function; input variables; Accuracy; Function approximation; Fuzzy sets; Fuzzy systems; Input variables; Neurons; Training; Fuzzy Systems; HHFNN; Neural networks; Takagi-Sugeno; Training algorithm;
Conference_Titel :
Information Engineering (ICIE), 2010 WASE International Conference on
Conference_Location :
Beidaihe, Hebei
Print_ISBN :
978-1-4244-7506-3
Electronic_ISBN :
978-1-4244-7507-0
DOI :
10.1109/ICIE.2010.11