Incremental Line Tangent Space Alignment Algorithm

Abstract

In this paper an incremental version of line tangent space alignment (LTSA) is proposed on the basis of incremental locally linear embedding (LLE) generalizations and the subsequent incremental Hessian locally linear embedding (HLLE). The main goal of this algorithm is to reduce the dimensionality of high-dimension manifolds into a lower dimension representation such that the significant characteristics of the dataset are preserved while adapting to newly added points arriving to the dataset. Experimental results are performed to verify how the new projection of points, along with the additional points, produces a good fit to the original manifold.