Unsupervised learning: Self-aggregation in scaled principal component space

Abstract

We demonstrate that data clustering amounts to a dynamic process of self-aggregation in which data objects move towards each other to form clusters, revealing the inherent pattern of similarity. Selfaggregation is governed by connectivity and occurs in a space obtained by a nonlinear scaling of principal component analysis (PCA). The method combines dimensionality reduction with clustering into a single framework. It can apply to both square similarity matrices and rectangular association matrices. © 2002 Springer-Verlag Berlin Heidelberg.