Publication detail

Non-Negative Tensor Factorization Accelerated Using GPGPU

ANTIKAINEN, J. HAVEL, J. JOŠTH, R. HEROUT, A. ZEMČÍK, P. HAUTA-KASARI, M.

Original Title

Non-Negative Tensor Factorization Accelerated Using GPGPU

English Title

Non-Negative Tensor Factorization Accelerated Using GPGPU

Type

journal article - other

Language

en

Original Abstract

This article presents an optimized algorithm for Non-Negative Tensor Factorization (NTF), implemented in the CUDA (Compute Uniform Device Architecture) framework, that runs on contemporary graphics processors and exploits their massive parallelism. The NTF implementation is primarily targeted for analysis of high-dimensional spectral images, including dimensionality reduction, feature extraction, and other tasks related to spectral imaging; however, the algorithm and its implementation are not limited to spectral imaging. The speed-ups measured on real spectral images are around 60-100x compared to a traditional  C implementation compiled with an optimizing compiler.  Since common problems in the field of spectral imaging may take hours on a state-of-the-art CPU, the speed-up achieved using a graphics card is attractive.  The implementation is publicly available in the form of a dynamically linked library, including an interface to MATLAB, and thus may be of help to researchers and engineers using NTF on large problems.

English abstract

This article presents an optimized algorithm for Non-Negative Tensor Factorization (NTF), implemented in the CUDA (Compute Uniform Device Architecture) framework, that runs on contemporary graphics processors and exploits their massive parallelism. The NTF implementation is primarily targeted for analysis of high-dimensional spectral images, including dimensionality reduction, feature extraction, and other tasks related to spectral imaging; however, the algorithm and its implementation are not limited to spectral imaging. The speed-ups measured on real spectral images are around 60-100x compared to a traditional  C implementation compiled with an optimizing compiler.  Since common problems in the field of spectral imaging may take hours on a state-of-the-art CPU, the speed-up achieved using a graphics card is attractive.  The implementation is publicly available in the form of a dynamically linked library, including an interface to MATLAB, and thus may be of help to researchers and engineers using NTF on large problems.

Keywords

Non-negative tensor factorization, spectral analysis, GPU

RIV year

2011

Released

14.03.2011

Publisher

NEUVEDEN

Location

NEUVEDEN

ISBN

1045-9219

Periodical

IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS

Year of study

2011

Number

1111

State

US

Pages count

7

Documents

BibTex


@article{BUT50517,
  author="Jukka {Antikainen} and Jiří {Havel} and Radovan {Jošth} and Adam {Herout} and Pavel {Zemčík} and Markku {Hauta-Kasari}",
  title="Non-Negative Tensor Factorization Accelerated Using GPGPU",
  annote="This article presents an optimized algorithm for Non-Negative Tensor
Factorization (NTF), implemented in the CUDA (Compute Uniform Device
Architecture) framework, that runs on contemporary graphics processors and
exploits their massive parallelism. The NTF implementation is primarily targeted
for analysis of high-dimensional spectral images, including dimensionality
reduction, feature extraction, and other tasks related to spectral imaging;
however, the algorithm and its implementation are not limited to spectral
imaging. The speed-ups measured on real spectral images are around 60-100x
compared to a traditional  C implementation compiled with an optimizing
compiler.  Since common problems in the field of spectral imaging may take hours
on a state-of-the-art CPU, the speed-up achieved using a graphics card is
attractive.  The implementation is publicly available in the form of
a dynamically linked library, including an interface to MATLAB, and thus may be
of help to researchers and engineers using NTF on large problems.",
  address="NEUVEDEN",
  chapter="50517",
  edition="NEUVEDEN",
  howpublished="print",
  institution="NEUVEDEN",
  journal="IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS",
  number="1111",
  volume="2011",
  year="2011",
  month="march",
  pages="0--0",
  publisher="NEUVEDEN",
  type="journal article - other"
}