A nonparametric data mapping technique for active initialization of the multilayer perceptron

Abstract

A new nonparametric feature mapping technique for pattern classification is proposed and compared experimentally with a principal component and Sammon’s mapping methods. We use the mapped training-set vectors for an active weights initialization of the multilayer perceptron classifier in a two-variate mapped space. Simulations have shown a usefulness of the proposed weights initialization method for designing the perceptrons when we need to obtain highly nonlinear decision boundaries.