UNIGE document Scientific Article
previous document  unige:97526  next document
add to browser collection

Employing GPU architectures for permutation-based indexing

Kruliš, Martin
Osipyan, Hasmik
Published in Multimedia Tools and Applications. 2017, vol. 76, no. 9, p. 11859-11887
Abstract Permutation-based indexing is one of the most popular techniques for the approximate nearest-neighbor search problem in high-dimensional spaces. Due to the exponential increase of multimedia data, the time required to index this data has become a serious constraint. One of the possible steps towards faster index construction is utilization of massively parallel platforms such as the GPGPU architectures. In this paper, we have analyzed the computational costs of individual steps of the permutation-based index construction in a high-dimensional feature space and summarized our hybrid CPU-GPU solution. Our experience gained from this research may be utilized in other individual problems that require computing Lp distances in high-dimensional spaces, parallel top-k selection, or partial sorting of multiple smaller sets. We also provide guidelines how to balance workload in hybrid CPU-GPU systems.
Keywords GPUParallelPermutation-based indexingApproximate similarity searchBitonic sorting
Full text
Article (Published version) (2.1 MB) - document accessible for UNIGE members only Limited access to UNIGE
Other version: http://link.springer.com/10.1007/s11042-016-3677-7
Research group Viper group
(ISO format)
KRULIŠ, Martin, OSIPYAN, Hasmik, MARCHAND-MAILLET, Stéphane. Employing GPU architectures for permutation-based indexing. In: Multimedia Tools and Applications, 2017, vol. 76, n° 9, p. 11859-11887. doi: 10.1007/s11042-016-3677-7 https://archive-ouverte.unige.ch/unige:97526

353 hits

1 download


Deposited on : 2017-10-10

Export document
Format :
Citation style :