Bottom-up nearest neighbor search for R-trees

Moon Bae Song, Kwang Jin Park, Ki S. Kong, Sang-Geun Lee

Research output: Contribution to journalArticle

2 Citations (Scopus)

Abstract

A new nearest neighbor (NN) search algorithm for R-trees was proposed that performed in a bottom-up manner. The proposed approach starts at the leaf level and traverses to the Root node. The experimental results demonstrate that the proposed NN search algorithm outperforms existing NN search algorithms which are based on the R-tree including the existing I/O optimal algorithm. The conventional R-trees were required to be modified to make the bottom-up approach possible. Two new pruning heuristics were also proposed where the first is the remnant property and the second is safe MBR property. The possible usage of the safe MBRs is to be investigated without the additional hash structure. It was found that the proposed approach outperforms existing approaches such as DFS and BFS without respect to the number of nearest neighbors.

Original languageEnglish
Pages (from-to)78-85
Number of pages8
JournalInformation Processing Letters
Volume101
Issue number2
DOIs
Publication statusPublished - 2007 Jan 31

Fingerprint

Nearest Neighbor Search
R-tree
Trees (mathematics)
Bottom-up
Search Algorithm
Pruning
Optimal Algorithm
Nearest Neighbor
Leaves
Roots
Heuristics
Experimental Results
Vertex of a graph
Demonstrate
Nearest neighbor search

Keywords

  • Bottom-up search
  • Databases
  • Nearest neighbor search
  • R-trees

ASJC Scopus subject areas

  • Computational Theory and Mathematics

Cite this

Bottom-up nearest neighbor search for R-trees. / Song, Moon Bae; Park, Kwang Jin; Kong, Ki S.; Lee, Sang-Geun.

In: Information Processing Letters, Vol. 101, No. 2, 31.01.2007, p. 78-85.

Research output: Contribution to journalArticle

Song, Moon Bae ; Park, Kwang Jin ; Kong, Ki S. ; Lee, Sang-Geun. / Bottom-up nearest neighbor search for R-trees. In: Information Processing Letters. 2007 ; Vol. 101, No. 2. pp. 78-85.
@article{833589e9432a4b0da8b1bb57e11d4bb7,
title = "Bottom-up nearest neighbor search for R-trees",
abstract = "A new nearest neighbor (NN) search algorithm for R-trees was proposed that performed in a bottom-up manner. The proposed approach starts at the leaf level and traverses to the Root node. The experimental results demonstrate that the proposed NN search algorithm outperforms existing NN search algorithms which are based on the R-tree including the existing I/O optimal algorithm. The conventional R-trees were required to be modified to make the bottom-up approach possible. Two new pruning heuristics were also proposed where the first is the remnant property and the second is safe MBR property. The possible usage of the safe MBRs is to be investigated without the additional hash structure. It was found that the proposed approach outperforms existing approaches such as DFS and BFS without respect to the number of nearest neighbors.",
keywords = "Bottom-up search, Databases, Nearest neighbor search, R-trees",
author = "Song, {Moon Bae} and Park, {Kwang Jin} and Kong, {Ki S.} and Sang-Geun Lee",
year = "2007",
month = "1",
day = "31",
doi = "10.1016/j.ipl.2006.08.005",
language = "English",
volume = "101",
pages = "78--85",
journal = "Information Processing Letters",
issn = "0020-0190",
publisher = "Elsevier",
number = "2",

}

TY - JOUR

T1 - Bottom-up nearest neighbor search for R-trees

AU - Song, Moon Bae

AU - Park, Kwang Jin

AU - Kong, Ki S.

AU - Lee, Sang-Geun

PY - 2007/1/31

Y1 - 2007/1/31

N2 - A new nearest neighbor (NN) search algorithm for R-trees was proposed that performed in a bottom-up manner. The proposed approach starts at the leaf level and traverses to the Root node. The experimental results demonstrate that the proposed NN search algorithm outperforms existing NN search algorithms which are based on the R-tree including the existing I/O optimal algorithm. The conventional R-trees were required to be modified to make the bottom-up approach possible. Two new pruning heuristics were also proposed where the first is the remnant property and the second is safe MBR property. The possible usage of the safe MBRs is to be investigated without the additional hash structure. It was found that the proposed approach outperforms existing approaches such as DFS and BFS without respect to the number of nearest neighbors.

AB - A new nearest neighbor (NN) search algorithm for R-trees was proposed that performed in a bottom-up manner. The proposed approach starts at the leaf level and traverses to the Root node. The experimental results demonstrate that the proposed NN search algorithm outperforms existing NN search algorithms which are based on the R-tree including the existing I/O optimal algorithm. The conventional R-trees were required to be modified to make the bottom-up approach possible. Two new pruning heuristics were also proposed where the first is the remnant property and the second is safe MBR property. The possible usage of the safe MBRs is to be investigated without the additional hash structure. It was found that the proposed approach outperforms existing approaches such as DFS and BFS without respect to the number of nearest neighbors.

KW - Bottom-up search

KW - Databases

KW - Nearest neighbor search

KW - R-trees

UR - http://www.scopus.com/inward/record.url?scp=33751268158&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33751268158&partnerID=8YFLogxK

U2 - 10.1016/j.ipl.2006.08.005

DO - 10.1016/j.ipl.2006.08.005

M3 - Article

AN - SCOPUS:33751268158

VL - 101

SP - 78

EP - 85

JO - Information Processing Letters

JF - Information Processing Letters

SN - 0020-0190

IS - 2

ER -