Access Restriction

Author Kosco, B.
Source IEEE Xplore Digital Library
Content type Text
Publisher Institute of Electrical and Electronics Engineers, Inc. (IEEE)
File Format PDF
Copyright Year ©1995
Language English
Subject Domain (in DDC) Computer science, information & general works ♦ Special computer methods
Subject Keyword Fuzzy systems ♦ Throughput ♦ Fuzzy sets ♦ Cancer ♦ Signal processing ♦ Image processing ♦ Arithmetic ♦ Statistics ♦ Adaptive systems
Abstract The paper shows how to combine any number of fuzzy systems and compares this new method with a weighted arithmetic mean. The new method combines the throughputs of the fuzzy systems and thus combines before it defuzzifies. The mean combines just the defuzzified outputs. The new method adds the output fuzzy set of each fuzzy system and then defuzzifies this sum in a higher-level additive fuzzy system. The combined fuzzy systems need not be additive fuzzy systems. The paper derives the exact form of the additive combiner. It reduces to an unweighted mean of centroids when all fuzzy sets have the same volume and all systems either have the same credibility weights or have binary credibility weights. The weighted mean ignores the inherent weighting information in the relative volumes. Its constant normalizer leads to scale defects when it combines weighted fuzzy systems even when all the weights are the same. The paper derives the first-order and second-order statistics of the general additive system and shows how fuzzy systems act as adaptive model-free statistical estimators.<<ETX>>
Description Author affiliation: Dept. of Electr. Eng., Univ. of Southern California, Los Angeles, CA, USA (Kosco, B.)
ISBN 0780324617
Educational Role Student ♦ Teacher
Age Range above 22 year
Educational Use Research ♦ Reading
Education Level UG and PG
Learning Resource Type Article
Publisher Date 1995-03-20
Rights Holder Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Size (in Bytes) 404.06 kB
Page Count 9
Starting Page 1855
Ending Page 1863

Source: IEEE Xplore Digital Library