Dynamic selection of suitable pivots for similarity search in metric spaces

Abstract

This paper presents a data structure based on Sparse Spatial Selection (SSS) for similarity searching. An algorithm that tries periodically to adjust pivots to the use of database index is presented. This index is dynamic. In this way, it is possible to improve the amount of discriminations done by the pivots. So, the primary objective of indexes is achieved: to reduce the number of distance function evaluations, as it is showed in the experimentation