Hybrid (CPU/GPU) Exact Nearest Neighbors Search in High-Dimensional Spaces

Research output: Chapter in Book/Report/Conference proceedingsConference contributionpeer-review

Abstract

In this paper, we propose a hybrid algorithm for exact nearest neighbors queries in high-dimensional spaces. Indexing structures typically used for exact nearest neighbors search become less efficient in high-dimensional spaces, effectively requiring brute-force search. Our method uses a massively-parallel approach to brute-force search that efficiently splits the computational load between CPU and GPU. We show that the performance of our algorithm scales linearly with the dimensionality of the data, improving upon previous approaches for high-dimensional datasets. The algorithm is implemented in Julia, a high-level programming language for numerical and scientific computing. It is openly available at https://github.com/davnn/ParallelNeighbors.jl.

Original languageEnglish
Title of host publicationArtificial Intelligence Applications and Innovations - 18th IFIP WG 12.5 International Conference, AIAI 2022, Proceedings
EditorsIlias Maglogiannis, Lazaros Iliadis, John Macintyre, Paulo Cortez
PublisherSpringer
Pages112-123
Number of pages12
ISBN (Print)9783031083365
DOIs
Publication statusPublished - 2022
Event18th IFIP WG 12.5 International Conference on Artificial Intelligence Applications and Innovations, AIAI 2022 - Hersonissos, Greece
Duration: 17 Jun 202220 Jun 2022

Publication series

NameIFIP Advances in Information and Communication Technology
Volume647 IFIP
ISSN (Print)1868-4238
ISSN (Electronic)1868-422X

Conference

Conference18th IFIP WG 12.5 International Conference on Artificial Intelligence Applications and Innovations, AIAI 2022
Country/TerritoryGreece
CityHersonissos
Period17.06.202220.06.2022

Keywords

  • CPU
  • Exact
  • GPU
  • Hybrid
  • k-NN
  • Nearest neighbors

Fingerprint

Dive into the research topics of 'Hybrid (CPU/GPU) Exact Nearest Neighbors Search in High-Dimensional Spaces'. Together they form a unique fingerprint.

Cite this