Laplacian Eigenmaps and spectral techniques for graph embedding

Abstract

Our focus is on enhancing the methodologies for dimensional reduction and clustering recognizing that traditional methods of ten fall short due to their inability to full consider the differential manifold structure underlying the data