Batch Support Vector Machine-Trained Fuzzy Classifier with channel equalization application

Abstract

This paper proposes a Batch Support Vector Machine-Trained Fuzzy Classifier (BSVM-FC). The BSVM-FC is a fuzzy system that consists of Takagi-Sugeno (TS)-type fuzzy rules. For structure learning of the BSVM-FC, there are no fuzzy rules initially. The BSVM-FC online generates all rules according to distributions of training data. A linear support vector machine (SVM) is used to tune the rule consequent parameters. The use of SVM is to give the classifier better generalization performance. Simulation is conducted to very the performance of the BSVM-FC. The BSVM-FC is applied to channel equalization. Comparisons with Gaussian-kernel SVM demonstrate that the BSVM-FC helps to speed up training and test times, and reduce classifier size without deteriorating the generalization ability.