Nonlinear Extension of the IRLS Classifier Using Clustering with Pairs of Prototypes
Classification is the primal problem in pattern recognition. The quadratic loss function is not good approximation of the misclassification error. In the presented paper a nonlinear extension of the IRLS classifier, which uses different loss functions, is proposed. The extension is done by means of fuzzy if-then rules. The fuzzy clustering with pairs of prototypes is applied to establish rules parameters values. Classification quality and computing time obtained for six benchmark databases is compared with the Lagrangian SVM method.
KeywordsFuzzy Clustering Rule-Base Classifiers IRLS
Unable to display preview. Download preview PDF.