A fast learning algorithm for uninorm-based fuzzy neural networks

This paper suggests a fast learning algorithm for weighted uninorm-based neural networks. Fuzzy neural networks are models capable to approximate functions with high accuracy and to generate transparent models through extraction of linguistic information from the resulting topology. A fuzzy neural n...

Full description

Saved in:
Bibliographic Details
Published in2012 Annual Meeting of the North American Fuzzy Information Processing Society (NAFIPS) pp. 1 - 6
Main Authors Lemos, A. P., Caminhas, W., Gomide, F.
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.08.2012
Subjects
Clustering algorithms
Fuzzy neural networks
Load forecasting
Load modeling
Network topology
Neurons
Topology
Online AccessGet full text

Cover

More Information
Summary:This paper suggests a fast learning algorithm for weighted uninorm-based neural networks. Fuzzy neural networks are models capable to approximate functions with high accuracy and to generate transparent models through extraction of linguistic information from the resulting topology. A fuzzy neural network model based on weighted uninorms has been developed recently. It was shown that this model approximates any continuous real function on a compact subset. In this paper we introduce a fast learning algorithm for this class of fuzzy neural networks based on ideas from extreme learning machine. The algorithm is detailed and computational experiments reported to illustrate the accuracy and time efficiency of the learning approach. The results show that neural fuzzy model is accurate and learning speed is as good as or faster than alternative neural network models.
ISBN:9781467323369
1467323365
DOI:10.1109/NAFIPS.2012.6290979