Ho--Kashyap classifier with generalization control

This paper introduces a new classifier design method based on a modification of the classical Ho-Kashyap procedure. The proposed method uses an absolute and squared approximation of the misclassification error rate to design a linear classifier. Moreover, the ability to generalize can be easily controlled and robustness of outliers is obtained. The proposed method of a classifier design maximizes the separation margin similarly as the support vector machine. In this paper, nine public domain benchmark datasets are used to evaluate the performance of the modified Ho-Kashyap classifier. A comparison with the support vector machine, kernel Fisher discriminant, regularized AdaBoost and radial-basis neural network is made. Large-scale simulations demonstrate the competitiveness of the proposed method with respect to the state-of-the-art classifiers.