A new approach for similarity queries using neighborhood graphs

Alexander OcsaCarlos BedregaErnesto Cuadros-Vargas

Proximity graphs have been widely used in the area of computational geometry; through a vicinity concept these graphs establish relations of similarity between elements of a set. In this paper we propose the use of Relative Neighborhood Graph (RNG) in metric spaces in order to eficiently answer similarity queries. Additionally we introduce a new algorithm for range queries and nearest neighbor queries making use of the spatial approximation in graphs. Experiments show that our proposal has a comparable performance in terms of Number of Distance Calculations (NDC) and time.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: