K-T.R.A.C.E: A kernel k-means procedure for classification

In a computational context, classification refers to assigning objects to different classes with respect to their features, which can be mapped to qualitative or quantitative variables. Several techniques have been developed recently to map the available information into a set of features (feature space) that improve the classification performance. Kernel functions provide a nonlinear mapping that implicitly transforms the input space to a new feature space where data can be separated, clustered and classified more easily. In this paper a kernel revised version of the Total Recognition by Adaptive Classification Experiments (T.R.A.C.E) algorithm, an iterative k-means like classification algorithm is presented.