Главная страница
russian   english
16+
<< back

Title of Article

DATA STRUCTURE WITH SMALL WORLD PROPERTIES FOR SOLVING NEAREST NEIGHBOR SEARCH PROBLEM IN METRIC SPACE


Issue
5
Date
2012

Article type
scientific article
UDC
004.057.4
Pages
409-415
Keywords
 


Authors
Ponomarenko Aleksandr Aleksandrovich
OOO «MeraLabs», Nizhniy Novgorod

Malkov Yuriy Andreevich
OOO «MeraLabs», Nizhniy Novgorod

Logvinov Andrey Aleksandrovich
OOO «MeraLabs», Nizhniy Novgorod

Krylov Vladimir Vladimirovich
OOO «MeraLabs», Nizhniy Novgorod


Abstract
A novel approach to solving the nearest neighbor search problem in metric space is considered. It is proposed as a data structure to use a graph with navigable small world properties and a gradient descent algorithm as a search algorithm. The problem of the existence of local minima is solved by a series of independent searches. Experimental data are presented to confirm logarithmic complexity of the search algorithm.

File (in Russian)