Negative correlation learning of neuro-fuzzy system ensembles

  • Authors:

  • Marcin Korytkowski;Rafał Scherer

  • Affiliations:

  • Department of Computer Engineering, Czestochowa University of Technology, Czestochowa, Poland and Olsztyn Academy of Computer Science and Management, Olsztyn, Poland;Department of Computer Engineering, Czestochowa University of Technology, Czestochowa, Poland and Academy of Management, SWSPiZ, Institute of Information Technology, Łódź, Poland

  • Venue:
  • ICAISC'10 Proceedings of the 10th international conference on Artificial intelligence and soft computing: Part I
  • Year:
  • 2010

Quantified Score

Hi-index 0.00

Visualization

Abstract

Ensembles of classifiers are sets of machine learning systems trained for the same task. The outputs of the systems are combined by various methods to obtain the classification result. Ensembles are proven to perform better than member weak learners. There are many methods for creating the ensembles. Most popular are Bagging and Boosting. In the paper we use the negative correlation learning to create an ensemble of Mamdani-type neuro-fuzzy systems. Negative correlation learning is a method which tries to decorrelate particular classifiers and to keep accuracy as high as possible. Neuro-fuzzy systems are good candidates for classification and machine learning problems as the knowledge is stored in the form of the fuzzy rules. The rules are relatively easy to create and interpret for humans, unlike in the case of other learning paradigms e.g. neural networks.