Approximate Nearest Neighbors in Limited Space
Abstract
We consider the -approximate nearest neighbor search problem: given a set of points in a -dimensional space, build a data structure that, given any query point , finds a point whose distance to is at most for an accuracy parameter . Our main result is a data structure that occupies only bits of space, assuming all point coordinates are integers in the range , i.e., the coordinates have bits of precision. This improves over the best previously known space bound of , obtained via the randomized dimensionality reduction method of Johnson and Lindenstrauss (1984). We also consider the more general problem of estimating all distances from a collection of query points to all data points , and provide almost tight upper and lower bounds for the space complexity of this problem.
Cite
@article{arxiv.1807.00112,
title = {Approximate Nearest Neighbors in Limited Space},
author = {Piotr Indyk and Tal Wagner},
journal= {arXiv preprint arXiv:1807.00112},
year = {2018}
}
Comments
COLT 2018