Euclidean lattices for high dimensional indexing and searching - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Reports (Research Report) Year : 2008

Euclidean lattices for high dimensional indexing and searching

Loïc Paulevé

Abstract

For similarity based searching, multimedia data are represented by one or more numerical vectors: we search the nearest neighbors of the query. Because of the huge number of these data and their high dimension, classical indexing technics are inefficient. The goal of this internship is to study the use of euclidean lattices for database indexing. Lattices have nice properties: they are spatial quantizers, thereby generate a partition of the space and decoding (quantization step) may be done very quickly. Then, we hope to be able to rapidly find a small space region containing data similar to a given query point, without reading all the database.
Fichier principal
Vignette du fichier
PI-1903.pdf (896.16 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

inria-00326262 , version 1 (02-10-2008)

Identifiers

  • HAL Id : inria-00326262 , version 1

Cite

Loïc Paulevé. Euclidean lattices for high dimensional indexing and searching. [Research Report] PI 1903, 2008, pp.57. ⟨inria-00326262⟩
129 View
103 Download

Share

Gmail Facebook Twitter LinkedIn More