Nonnegative matrix and tensor factorizations: applications to exploratory multi-way data analysis and blind source separation

This book provides an overview of existing models, methods and algorithms as an introduction to the topic, and presents several new techniques and approaches developed by the authors that can be applied to non-negative matrix and tensor factorizations. The authors focus on algorithms which are fast and robust in the field of NMFs, as well as related models which have much flexibility, and therefore are the most useful in practice. Using generalized cost functionssuch as Bregman, alpha and beta divergences, the authors present practical implementations of several types of robust algorithms, in particular Multiplicative, Projected Gradient and Quasi Newton algorithms. A comparative analysis ofthe different methods is given in order to identify approximation error and complexity

• ISBN: 978-0-470-74666-0
• Editorial: John Wiley & Sons
• Encuadernacion: Cartoné
• Páginas: 504
• Fecha Publicación: 11/09/2009
• Nº Volúmenes: 1
• Idioma: Inglés