Linear-time Training of Nonlinear Low-Dimensional Embeddings
Max Vladymyrov, Miguel Á. Carreira-Perpiñán · 2014
Introduction. Dimensionality reduction is an important task in machine learning. It arrises when there is a need for exploratory analysis of a dataset, to reveal hidden structure of the data, or as a pre-processing step, by extracting low-dimensional features that are useful for nearest-neighbor retrieval, classification, search or other applications, in an unsupervised way. We focus on a well-known class of dimensionality reduction algorithms, called embedding algorithms based on pairwise affinities.