Skip to main navigation Skip to search Skip to main content

Redundancy-Controlled Feature Selection for Fuzzy Neural Networks

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Group Lasso term has the group compression role on the value of the corresponding variables. In the fuzzy rule based system, for realizing fuzzification operation each feature corresponds to different number of membership functions. Inspired by the Group Lasso idea, the width of Gaussian membership function based Group Lasso regularization is designed to realize feature selection in the multiple inputs multiple outputs (MIMO) fuzzy neural network (FNN). To take into account the linear/nonlinear redundancy between features. Using the Pearson's correlation coefficient, a redundancy-controlled regularizer is also added into the traditional error function to further improve the performance of feature selection, i.e., control redundancy between features. We demonstrate the effectiveness of the strategy on three benchmark classification data sets.

Original languageEnglish
Title of host publication2021 International Conference on High Performance Big Data and Intelligent Systems, HPBD and IS 2021
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages172-176
Number of pages5
ISBN (Electronic)9781665413275
DOIs
StatePublished - 2021
Event2021 International Conference on High Performance Big Data and Intelligent Systems, HPBD and IS 2021 - Macau, China
Duration: 5 Dec 20217 Dec 2021

Publication series

Name2021 International Conference on High Performance Big Data and Intelligent Systems, HPBD and IS 2021

Conference

Conference2021 International Conference on High Performance Big Data and Intelligent Systems, HPBD and IS 2021
Country/TerritoryChina
CityMacau
Period5/12/217/12/21

Keywords

  • Group Lasso
  • Pearson
  • feature redundancy
  • fuzzy neural network

Fingerprint

Dive into the research topics of 'Redundancy-Controlled Feature Selection for Fuzzy Neural Networks'. Together they form a unique fingerprint.

Cite this