Dirichlet densifiers for improved commute times estimation

Empreu sempre aquest identificador per citar o enllaçar aquest ítem http://hdl.handle.net/10045/88752
Información del item - Informació de l'item - Item information
Títol: Dirichlet densifiers for improved commute times estimation
Autors: Curado, Manuel | Escolano, Francisco | Lozano, Miguel Angel | Hancock, Edwin R.
Grups d'investigació o GITE: Laboratorio de Investigación en Visión Móvil (MVRLab)
Centre, Departament o Servei: Universidad de Alicante. Departamento de Ciencia de la Computación e Inteligencia Artificial
Paraules clau: Graph densification | Dirichlet problems | Random walkers | Commute times
Àrees de coneixement: Ciencia de la Computación e Inteligencia Artificial
Data de publicació: de juliol-2019
Editor: Elsevier
Citació bibliogràfica: Pattern Recognition. 2019, 91: 56-68. doi:10.1016/j.patcog.2019.02.012
Resum: In this paper, we develop a novel Dirichlet densifier that can be used to increase the edge density in undirected graphs. Dirichlet densifiers are implicit minimizers of the spectral gap for the Laplacian spectrum of a graph. One consequence of this property is that they can be used improve the estimation of meaningful commute distances for mid-size graphs by means of topological modifications of the original graphs. This results in a better performance in clustering and ranking. To do this, we identify the strongest edges and from them construct the so called line graph, where the nodes are the potential q −step reachable edges in the original graph. These strongest edges are assumed to be stable. By simulating random walks on the line graph, we identify potential new edges in the original graph. This approach is fully unsupervised and it is both more scalable and robust than recent explicit spectral methods, such as the Semi-Definite Programming (SDP) densifier and the sufficient condition for decreasing the spectral gap. Experiments show that our method is only outperformed by some choices of the parameters of a related method, the anchor graph, which relies on pre-computing clusters representatives, and that the proposed method is effective on a variety of real-world datasets.
Patrocinadors: M. Curado, F. Escolano and M.A. Lozano are funded by the projects TIN2015-69077-P and BES2013-064482 of the Spanish Government.
URI: http://hdl.handle.net/10045/88752
ISSN: 0031-3203 (Print) | 1873-5142 (Online)
DOI: 10.1016/j.patcog.2019.02.012
Idioma: eng
Tipus: info:eu-repo/semantics/article
Drets: © 2019 Elsevier Ltd.
Revisió científica: si
Versió de l'editor: https://doi.org/10.1016/j.patcog.2019.02.012
Apareix a la col·lecció: INV - MVRLab - Artículos de Revistas

Arxius per aquest ítem:
Arxius per aquest ítem:
Arxiu Descripció Tamany Format  
Thumbnail2019_Curado_etal_PatternRecogn_final.pdfVersión final (acceso restringido)2,32 MBAdobe PDFObrir     Sol·licitar una còpia
Thumbnail2019_Curado_etal_PatternRecogn_preprint.pdfPreprint (acceso abierto)3,46 MBAdobe PDFObrir Vista prèvia


Tots els documents dipositats a RUA estan protegits per drets d'autors. Alguns drets reservats.