Considerations on parallelizing nonnegative matrix factorization for hyperspectral data unmixing

Stefan Robila, Lukasz G. MacIak

Research output: Contribution to journalArticleResearchpeer-review

24 Citations (Scopus)

Abstract

Nonnegative matrix factorization (NMF) is a recently developed linear unmixing technique that assumes that the original sources and transform were positively defined. Given that the linear mixing model (LMM) for hyperspectral data requires positive endmembers and abundances, with only minor modifications, NMF can be used to solve LMM. Traditionally, NMF solutions include an iterative process resulting in considerable execution times. In this letter, we provide two novel algorithms aimed at speeding the NMF through parallel processing: the first based on the traditional multiplicative solution and the second modifying an adaptive projected gradient technique known to provide better convergence. The algorithms' implementations were tested on various data sets; the results suggest that a significant speedup can be achieved without decrease in accuracy. This supports the further use of NMF for linear unmixing.

Original languageEnglish
Article number4674612
Pages (from-to)57-61
Number of pages5
JournalIEEE Geoscience and Remote Sensing Letters
Volume6
Issue number1
DOIs
StatePublished - 1 Jan 2009

Fingerprint

Factorization
matrix
transform
Processing

Keywords

  • Hyperspectral data
  • Linear algorithms
  • Linear unmixing
  • Nonnegative matrix factorization (NMF)
  • Parallel processing
  • Remote sensing

Cite this

@article{5853debeabb44455a2d2b9e598e16bba,
title = "Considerations on parallelizing nonnegative matrix factorization for hyperspectral data unmixing",
abstract = "Nonnegative matrix factorization (NMF) is a recently developed linear unmixing technique that assumes that the original sources and transform were positively defined. Given that the linear mixing model (LMM) for hyperspectral data requires positive endmembers and abundances, with only minor modifications, NMF can be used to solve LMM. Traditionally, NMF solutions include an iterative process resulting in considerable execution times. In this letter, we provide two novel algorithms aimed at speeding the NMF through parallel processing: the first based on the traditional multiplicative solution and the second modifying an adaptive projected gradient technique known to provide better convergence. The algorithms' implementations were tested on various data sets; the results suggest that a significant speedup can be achieved without decrease in accuracy. This supports the further use of NMF for linear unmixing.",
keywords = "Hyperspectral data, Linear algorithms, Linear unmixing, Nonnegative matrix factorization (NMF), Parallel processing, Remote sensing",
author = "Stefan Robila and MacIak, {Lukasz G.}",
year = "2009",
month = "1",
day = "1",
doi = "10.1109/LGRS.2008.2005793",
language = "English",
volume = "6",
pages = "57--61",
journal = "IEEE Geoscience and Remote Sensing Letters",
issn = "1545-598X",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "1",

}

Considerations on parallelizing nonnegative matrix factorization for hyperspectral data unmixing. / Robila, Stefan; MacIak, Lukasz G.

In: IEEE Geoscience and Remote Sensing Letters, Vol. 6, No. 1, 4674612, 01.01.2009, p. 57-61.

Research output: Contribution to journalArticleResearchpeer-review

TY - JOUR

T1 - Considerations on parallelizing nonnegative matrix factorization for hyperspectral data unmixing

AU - Robila, Stefan

AU - MacIak, Lukasz G.

PY - 2009/1/1

Y1 - 2009/1/1

N2 - Nonnegative matrix factorization (NMF) is a recently developed linear unmixing technique that assumes that the original sources and transform were positively defined. Given that the linear mixing model (LMM) for hyperspectral data requires positive endmembers and abundances, with only minor modifications, NMF can be used to solve LMM. Traditionally, NMF solutions include an iterative process resulting in considerable execution times. In this letter, we provide two novel algorithms aimed at speeding the NMF through parallel processing: the first based on the traditional multiplicative solution and the second modifying an adaptive projected gradient technique known to provide better convergence. The algorithms' implementations were tested on various data sets; the results suggest that a significant speedup can be achieved without decrease in accuracy. This supports the further use of NMF for linear unmixing.

AB - Nonnegative matrix factorization (NMF) is a recently developed linear unmixing technique that assumes that the original sources and transform were positively defined. Given that the linear mixing model (LMM) for hyperspectral data requires positive endmembers and abundances, with only minor modifications, NMF can be used to solve LMM. Traditionally, NMF solutions include an iterative process resulting in considerable execution times. In this letter, we provide two novel algorithms aimed at speeding the NMF through parallel processing: the first based on the traditional multiplicative solution and the second modifying an adaptive projected gradient technique known to provide better convergence. The algorithms' implementations were tested on various data sets; the results suggest that a significant speedup can be achieved without decrease in accuracy. This supports the further use of NMF for linear unmixing.

KW - Hyperspectral data

KW - Linear algorithms

KW - Linear unmixing

KW - Nonnegative matrix factorization (NMF)

KW - Parallel processing

KW - Remote sensing

UR - http://www.scopus.com/inward/record.url?scp=58149520173&partnerID=8YFLogxK

U2 - 10.1109/LGRS.2008.2005793

DO - 10.1109/LGRS.2008.2005793

M3 - Article

VL - 6

SP - 57

EP - 61

JO - IEEE Geoscience and Remote Sensing Letters

JF - IEEE Geoscience and Remote Sensing Letters

SN - 1545-598X

IS - 1

M1 - 4674612

ER -