Comparative Study of Nonlinear Methods for Manifold Learning

Download insert_drive_file

Authors:

Bernstein A., Burnaev E., Erofeev P.

Journal / Event:

Proc. of the conf. "Information Technologies and Systems". 2012. P. 85–91.

Abstract:

In this paper manifold embedding and reconstruction procedures are considered in the scope of unsupervised dimension reduction problem. Standard approaches (Isomap, LLE, LTSA, etc.) are compared to newly proposed Grassman-Stiefel Eigenmaps (GSE) algorithm. It turned out that GSE provides best manifold reconstruction abilities on test problems.

Keywords: Data Analysis, Dimension Reduction