Assessment of the Effect of Noise on an Unsupervised Feature Selection Method for Generative Topographic Mapping

Unsupervised feature relevance determination and feature selection for dimensionality reduction are important issues in many clustering problems. An unsupervised feature selection method for general Finite Mixture Models was recently proposed and subsequently extended to Generative Topographic Mapping (GTM), a nonlinear manifold learning constrained mixture model for data clustering and visualization. Some of the results of a previous preliminary assessment of this method for GTM suggested that its performance may be affected by the presence of uninformative noise in the dataset. In this brief study, we test in some detail such limitation of the method.