English

Approximate Nearest Neighbors in Limited Space

Data Structures and Algorithms 2018-07-03 v1 Computational Geometry

Abstract

We consider the (1+ϵ)(1+\epsilon)-approximate nearest neighbor search problem: given a set XX of nn points in a dd-dimensional space, build a data structure that, given any query point yy, finds a point xXx \in X whose distance to yy is at most (1+ϵ)minxXxy(1+\epsilon) \min_{x \in X} \|x-y\| for an accuracy parameter ϵ(0,1)\epsilon \in (0,1). Our main result is a data structure that occupies only O(ϵ2nlog(n)log(1/ϵ))O(\epsilon^{-2} n \log(n) \log(1/\epsilon)) bits of space, assuming all point coordinates are integers in the range {nO(1)nO(1)}\{-n^{O(1)} \ldots n^{O(1)}\}, i.e., the coordinates have O(logn)O(\log n) bits of precision. This improves over the best previously known space bound of O(ϵ2nlog(n)2)O(\epsilon^{-2} n \log(n)^2), 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 XX, and provide almost tight upper and lower bounds for the space complexity of this problem.

Keywords

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

R2 v1 2026-06-23T02:46:44.092Z