Reverse Nearest Neighbors Search in High Dimensions using Locality-Sensitive Hashing
Abstract
We investigate the problem of finding reverse nearest neighbors efficiently. Although provably good solutions exist for this problem in low or fixed dimensions, to this date the methods proposed in high dimensions are mostly heuristic. We introduce a method that is both provably correct and efficient in all dimensions, based on a reduction of the problem to one instance of $\e$-nearest neighbor search plus a controlled number of instances of {\em exhaustive $r$-\pleb}, a variant of {\em Point Location among Equal Balls} where all the $r$-balls centered at the data points that contain the query point are sought for, not just one. The former problem has been extensively studied and elegantly solved in high dimensions using Locality-Sensitive Hashing (LSH) techniques. By contrast, the latter problem has a complexity that is still not fully understood. We revisit the analysis of the LSH scheme for exhaustive $r$-\pleb using a somewhat refined notion of locality-sensitive family of hash function, which brings out a meaningful output-sensitive term in the complexity of the problem. Our analysis, combined with a non-isometric lifting of the data, enables us to answer exhaustive $r$-\pleb queries (and down the road reverse nearest neighbors queries) efficiently. Along the way, we obtain a simple algorithm for answering exact nearest neighbor queries, whose complexity is parametrized by some {\em condition number} measuring the inherent difficulty of a given instance of the problem.
- Publication:
-
arXiv e-prints
- Pub Date:
- November 2010
- DOI:
- 10.48550/arXiv.1011.4955
- arXiv:
- arXiv:1011.4955
- Bibcode:
- 2010arXiv1011.4955A
- Keywords:
-
- Computer Science - Computational Geometry;
- Computer Science - Data Structures and Algorithms