Generalization of the K-SVD algorithm for minimization of ß-divergence

[EN] In this paper, we propose, describe, and test a modification of the K-SVD algorithm. Given a set of training data, the proposed algorithm computes an overcomplete dictionary by minimizing the ß-divergence () between the data and its representation as linear combinations of atoms of the dictiona...

Descripción completa

Detalles Bibliográficos
Autores: García Mollá, Víctor Manuel|||0000-0003-4768-7367, Vidal Maciá, Antonio Manuel, Alonso-Jordá, Pedro|||0000-0002-6882-6592, San Juan-Sebastian, Pablo, Virtanen, T.
Tipo de recurso: artículo
Fecha de publicación:2019
País:España
Institución:Universitat Politècnica de València (UPV)
Repositorio:RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Idioma:inglés
OAI Identifier:oai:riunet.upv.es:10251/123503
Acceso en línea:https://riunet.upv.es/handle/10251/123503
Access Level:acceso abierto
Palabra clave:K-SVD
Nonnegative K-SVD
Beta-divergence
NMF
Matching pursuit algorithms
CIENCIAS DE LA COMPUTACION E INTELIGENCIA ARTIFICIAL
Descripción
Sumario:[EN] In this paper, we propose, describe, and test a modification of the K-SVD algorithm. Given a set of training data, the proposed algorithm computes an overcomplete dictionary by minimizing the ß-divergence () between the data and its representation as linear combinations of atoms of the dictionary, under strict sparsity restrictions. For the special case , the proposed algorithm minimizes the Frobenius norm and, therefore, for the proposed algorithm is equivalent to the original K-SVD algorithm. We describe the modifications needed and discuss the possible shortcomings of the new algorithm. The algorithm is tested with random matrices and with an example based on speech separation.