Nonlinear Dataset Structure Detection Based on Subsets Clustering

Provided by: Binary Information Press
Topic: Data Management
Format: PDF
For resolving the detection problem of nonlinear distribution structure of dataset, a novel subsets clustering algorithm based on Riemannian geometry is proposed in this paper. Comparing with the existing manifold learning methods, the proposed algorithm does not focus on the preservation of high-dimension neighborhood information and reconstruction in low-dimension space, but detect and find the approximately linear distribution partitions of the high-dimension dataset, and output a simple model directly which can represent these linear partitions' information.

Find By Topic