Preliminary evaluation of batch-learning self-organizing map algorithm on a graphic processor

Akihiro Shitara, Yuri Nishikawa, Masato Yoshimi, Takashi Abe, Toshimichi Ikemura, Hideharu Amano

Research output: Chapter in Book/Report/Conference proceedingConference contribution

1 Citation (Scopus)

Abstract

In this paper, we introduce a GPU implementation and evaluation of batch learning self-organizing maps (BL-SOM) algorithm, which improves Kohonen's original SOM algorithm by making input data sequence independent from learning process. We used CUDA provided by NVIDIA Corporation for parallel programming, profiling, and data flow optimization so as to exploit inherent datalevel parallelism of the algorithm. With various parameter combinations, implementation on GTX280 achieved 250 times higher performance compared to Intel's Core2Quad Q6600 2.40GHz when parameters of map size, dimension of vectors, learning size and iteration of learning were 960×960, 136, 70 and 1, respectively.

Original languageEnglish
Title of host publicationProceedings of the 9th IASTED International Conference on Parallel and Distributed Computing and Networks, PDCN 2010
Pages96-104
Number of pages9
Publication statusPublished - 2010
Event9th IASTED International Conference on Parallel and Distributed Computing and Networks, PDCN 2010 - Innsbruck, Austria
Duration: 2010 Feb 162010 Feb 18

Other

Other9th IASTED International Conference on Parallel and Distributed Computing and Networks, PDCN 2010
CountryAustria
CityInnsbruck
Period10/2/1610/2/18

Fingerprint

Self organizing maps
Parallel programming
Industry

Keywords

  • CUDA
  • GPGPU
  • GPU
  • Self-organizing map

ASJC Scopus subject areas

  • Computational Theory and Mathematics
  • Computer Networks and Communications
  • Software

Cite this

Shitara, A., Nishikawa, Y., Yoshimi, M., Abe, T., Ikemura, T., & Amano, H. (2010). Preliminary evaluation of batch-learning self-organizing map algorithm on a graphic processor. In Proceedings of the 9th IASTED International Conference on Parallel and Distributed Computing and Networks, PDCN 2010 (pp. 96-104)

Preliminary evaluation of batch-learning self-organizing map algorithm on a graphic processor. / Shitara, Akihiro; Nishikawa, Yuri; Yoshimi, Masato; Abe, Takashi; Ikemura, Toshimichi; Amano, Hideharu.

Proceedings of the 9th IASTED International Conference on Parallel and Distributed Computing and Networks, PDCN 2010. 2010. p. 96-104.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Shitara, A, Nishikawa, Y, Yoshimi, M, Abe, T, Ikemura, T & Amano, H 2010, Preliminary evaluation of batch-learning self-organizing map algorithm on a graphic processor. in Proceedings of the 9th IASTED International Conference on Parallel and Distributed Computing and Networks, PDCN 2010. pp. 96-104, 9th IASTED International Conference on Parallel and Distributed Computing and Networks, PDCN 2010, Innsbruck, Austria, 10/2/16.
Shitara A, Nishikawa Y, Yoshimi M, Abe T, Ikemura T, Amano H. Preliminary evaluation of batch-learning self-organizing map algorithm on a graphic processor. In Proceedings of the 9th IASTED International Conference on Parallel and Distributed Computing and Networks, PDCN 2010. 2010. p. 96-104
Shitara, Akihiro ; Nishikawa, Yuri ; Yoshimi, Masato ; Abe, Takashi ; Ikemura, Toshimichi ; Amano, Hideharu. / Preliminary evaluation of batch-learning self-organizing map algorithm on a graphic processor. Proceedings of the 9th IASTED International Conference on Parallel and Distributed Computing and Networks, PDCN 2010. 2010. pp. 96-104
@inproceedings{e55725e07f0e4d2a9629f4b83039a375,
title = "Preliminary evaluation of batch-learning self-organizing map algorithm on a graphic processor",
abstract = "In this paper, we introduce a GPU implementation and evaluation of batch learning self-organizing maps (BL-SOM) algorithm, which improves Kohonen's original SOM algorithm by making input data sequence independent from learning process. We used CUDA provided by NVIDIA Corporation for parallel programming, profiling, and data flow optimization so as to exploit inherent datalevel parallelism of the algorithm. With various parameter combinations, implementation on GTX280 achieved 250 times higher performance compared to Intel's Core2Quad Q6600 2.40GHz when parameters of map size, dimension of vectors, learning size and iteration of learning were 960×960, 136, 70 and 1, respectively.",
keywords = "CUDA, GPGPU, GPU, Self-organizing map",
author = "Akihiro Shitara and Yuri Nishikawa and Masato Yoshimi and Takashi Abe and Toshimichi Ikemura and Hideharu Amano",
year = "2010",
language = "English",
isbn = "9780889868205",
pages = "96--104",
booktitle = "Proceedings of the 9th IASTED International Conference on Parallel and Distributed Computing and Networks, PDCN 2010",

}

TY - GEN

T1 - Preliminary evaluation of batch-learning self-organizing map algorithm on a graphic processor

AU - Shitara, Akihiro

AU - Nishikawa, Yuri

AU - Yoshimi, Masato

AU - Abe, Takashi

AU - Ikemura, Toshimichi

AU - Amano, Hideharu

PY - 2010

Y1 - 2010

N2 - In this paper, we introduce a GPU implementation and evaluation of batch learning self-organizing maps (BL-SOM) algorithm, which improves Kohonen's original SOM algorithm by making input data sequence independent from learning process. We used CUDA provided by NVIDIA Corporation for parallel programming, profiling, and data flow optimization so as to exploit inherent datalevel parallelism of the algorithm. With various parameter combinations, implementation on GTX280 achieved 250 times higher performance compared to Intel's Core2Quad Q6600 2.40GHz when parameters of map size, dimension of vectors, learning size and iteration of learning were 960×960, 136, 70 and 1, respectively.

AB - In this paper, we introduce a GPU implementation and evaluation of batch learning self-organizing maps (BL-SOM) algorithm, which improves Kohonen's original SOM algorithm by making input data sequence independent from learning process. We used CUDA provided by NVIDIA Corporation for parallel programming, profiling, and data flow optimization so as to exploit inherent datalevel parallelism of the algorithm. With various parameter combinations, implementation on GTX280 achieved 250 times higher performance compared to Intel's Core2Quad Q6600 2.40GHz when parameters of map size, dimension of vectors, learning size and iteration of learning were 960×960, 136, 70 and 1, respectively.

KW - CUDA

KW - GPGPU

KW - GPU

KW - Self-organizing map

UR - http://www.scopus.com/inward/record.url?scp=77954595546&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=77954595546&partnerID=8YFLogxK

M3 - Conference contribution

SN - 9780889868205

SP - 96

EP - 104

BT - Proceedings of the 9th IASTED International Conference on Parallel and Distributed Computing and Networks, PDCN 2010

ER -