Base Classifiers
Ludmila I. Kuncheva · 2004
Chapter 2 explains the base classifiers commonly used in multiple classifier systems. From the parametric classifier group we consider linear discriminant classifier (LDC), quadratic discriminant classifier (QDC) and regularized discriminant analysis. From the non-parametric classifier group we consider the multinomial classifier, kernel classifiers (Parzen) and prototype classifiers (k-nearest neighbor (k-nn), data editing for k-nn and distance-based k-nn variants). Decision-tree classifiers are introduced and algorithms for growing/pruning are presented. Multi-layer perceptron (MLP) neural network classifiers are explained. The standard backpropagation training algorithm for MLP is sketched.