High-Dimensional Access Methods for Efficient Similarity Queries

Retrieving similar complex documents such as images, sounds, DNA sequences, from within a large collection is an issues of main importance. While content modeling and retrieval algorithms tends to perform more and more efficiently, the methods to access the documents through their abstraction in form of high-dimensional feature vectors perform still poorly. In this report we detail the different access methods that have been proposed to perform similarity queries on multi-dimensional feature space, we present the reason of their inefficiency in high-dimensional feature space and finally we review the attempts to solve these issues