Hybrid LSH: Faster Near Neighbors Reporting in High-dimensional Space

TL;DR

This paper introduces a hybrid search method combining LSH and linear search for high-dimensional near neighbor problems, dynamically choosing the best strategy based on data distribution to improve efficiency.

Contribution

It proposes an auxiliary data structure that estimates search costs, enabling adaptive strategy selection in high-dimensional $r$-NN problems.

Findings

Hybrid approach outperforms pure LSH and linear search in experiments.

Efficient cost estimation guides strategy choice regardless of data distribution.

Method is compatible with recent LSH techniques.

Abstract

We study the $r$-near neighbors reporting problem ($r$-NN), i.e., reporting \emph{all} points in a high-dimensional point set $S$ that lie within a radius $r$ of a given query point $q$. Our approach builds upon on the locality-sensitive hashing (LSH) framework due to its appealing asymptotic sublinear query time for near neighbor search problems in high-dimensional space. A bottleneck of the traditional LSH scheme for solving $r$-NN is that its performance is sensitive to data and query-dependent parameters. On datasets whose data distributions have diverse local density patterns, LSH with inappropriate tuning parameters can sometimes be outperformed by a simple linear search. In this paper, we introduce a hybrid search strategy between LSH-based search and linear search for $r$-NN in high-dimensional space. By integrating an auxiliary data structure into LSH hash tables, we can efficiently estimate the computational cost of LSH-based search for a given query regardless of the data distribution. This means that we are able to choose the appropriate search strategy between LSH-based search and linear search to achieve better performance. Moreover, the integrated data structure is time efficient and fits well with many recent state-of-the-art LSH-based approaches. Our experiments on real-world datasets show that the hybrid search approach outperforms (or is comparable to) both LSH-based search and linear search for a wide range of search radii and data distributions in high-dimensional space.