Dissimilarity Representations Using l_p-norms in Eigen Spaces

摘要

This paper presents an empirical evaluation on a dissimilarity measure strategy by which dissimilarity-based classifications (DBC) can be implemented efficiently. In DBC, classification is not based on feature measurements of individual objects (a set of attributes), but rather on a suitable dissimilarity measure among the individual objects (pair-wise object comparisons). One problem of DBC is the high dimensionality of the dissimilarity space. To address this issue, two kinds of solutions have been proposed in the literature: prototype selection (PS)-based methods and dimension reduction (DR)-based methods. In this paper, instead of utilizing the PS-based or DR-based methods, we study a way of performing DBC in Eigen spaces (termed as EDBC), spanned by the subset of principal eigenvectors, extracted from the training dataset through a principal component analysis. Specifically, in EDBC, we use Ip-norms in combination with a rotation to eigenvectors to compute distances in a vector space, for constructing a dissimilarity-based classifier. The experimental results, obtained with artificial and real-life benchmark datasets, demonstrate that when the dimensionality of the Eigen spaces has been appropriately chosen, the classification accuracy of DBC can be improved.