Dimensionality Reduction with Unsupervised Nearest Neighbors

The growing information infrastructure in a variety of disciplines involves an increasing requirement for efficient data mining techniques. Fast dimensionality reduction methods are important for understanding and processing of large data sets of high-dimensional patterns. In this work, unsupervised nearest neighbors (UNN), an efficient iterative method for dimensionality reduction, is presented. Starting with an introduction to machine learning and dimensionality reduction, the framework for unsupervised regression is introduced, which is the basis of UNN. Algorithmic variants are developed step by step, reaching from a simple iterative strategy in discrete latent spaces to stochastic kernel-based submanifolds with independent parameterizations. Experimental comparisons to related methodologies taking into account realworld data sets and missing data scenarios show the behavior of UNN in practical scenarios.