Hypercube LSH for approximate near neighbors

Onderzoeksoutput: Hoofdstuk in Boek/Rapport/CongresprocedureConferentiebijdrageAcademicpeer review

2 Citaten (Scopus)

Samenvatting

A celebrated technique for finding near neighbors for the angular distance involves using a set of random hyperplanes to partition the space into hash regions [Charikar, STOC 2002]. Experiments later showed that using a set of orthogonal hyperplanes, thereby partitioning the space into the Voronoi regions induced by a hypercube, leads to even better results [Terasawa and Tanaka, WADS 2007]. However, no theoretical explanation for this improvement was ever given, and it remained unclear how the resulting hypercube hash method scales in high dimensions. In this work, we provide explicit asymptotics for the collision probabilities when using hypercubes to partition the space. For instance, two near-orthogonal vectors are expected to collide with probability ( 1/n)d+o(d) in dimension d, compared to (12)d when using random hyperplanes. Vectors at angle Φ3 collide with probability (√3)d+o(d), compared to (23 )d for random hyperplanes, and near-parallel vectors collide with similar asymptotic probabilities in both cases. For c-approximate nearest neighbor searching, this translates to a decrease in the exponent → of locality-sensitive hashing (LSH) methods of a factor up to log2(φ) ~ 1.652 compared to hyperplane LSH. For c = 2, we obtain → p = 0.302 + o(1) for hypercube LSH, improving upon the p = 0.377 for hyperplane LSH. We further describe how to use hypercube LSH in practice, and we consider an example application in the area of lattice algorithms.

Originele taal-2Engels
Titel42nd International Symposium on Mathematical Foundations of Computer Science, MFCS 2017
RedacteurenK.G. Larsen, H.L. Bodlaender, J.F. Raskin
Plaats van productieDagstuhl
UitgeverijSchloss Dagstuhl - Leibniz-Zentrum für Informatik
Pagina's7:1 - 7:20
ISBN van elektronische versie978-3-95977-046-0
DOI's
StatusGepubliceerd - 1 nov 2017
Extern gepubliceerdJa
Evenement42nd International Symposium on Mathematical Foundations of Computer Science (MFCS 2017) - Aalborg, Denemarken
Duur: 21 aug 201725 aug 2017
Congresnummer: 42
http://mfcs2017.cs.aau.dk/

Publicatie series

NaamLeibniz International Proceedings in Informatics (LIPIcs)
Volume83
ISSN van elektronische versie1868-8969

Congres

Congres42nd International Symposium on Mathematical Foundations of Computer Science (MFCS 2017)
Verkorte titelMFCS 2017
LandDenemarken
StadAalborg
Periode21/08/1725/08/17
Internet adres

Vingerafdruk Duik in de onderzoeksthema's van 'Hypercube LSH for approximate near neighbors'. Samen vormen ze een unieke vingerafdruk.

Citeer dit