Adaptive Neighborhood Graph for LTSA Learning Algorithm without FreeParameter

Abstract

Tangent Space Alignment (LTSA) algorithm is a classic local nonlinear manifold learning algorithm based on the information about local neighborhood space, i.e., local tangent space with respect to each point in dataset, which aims at finding the low-dimension intrinsic structure lie in high dimensional data space for the purpose of dimensionality reduction. In this paper, we present a novel learning algorithm, called 3N-LTSA which needs no free parameter in contrast to LTSA by using an adaptive nearest neighborhood graph. Experimental results show that 3N-LTSA algorithm without free parameter performs more practical and simple algorithm than LTSA.