Mixed double-multiple precision version of hyperplane constrained method for singular value decomposition

• Yadani Kenichi

  Graduate School of Informatics, Kyoto University

• Kondo Koichi

  Faculty of Science and Engineering, Doshisha University

• Iwasaki Masashi

  Department of Informatics and Environmental Science, Kyoto Prefectural University

Abstract

In this paper, we design a mixed double-multiple precision version of the hyperplane constrained method for singular value decomposition (SVD), which is based on solving nonlinear systems with the solutions constrained on hyperplanes. We also propose its hybrid method in order to shorten the running time. Through some numerical examples for matrices with small singular values, it is shown that, by new versions, the SVD is computable with high relative accuracy.

Journal

• JSIAM Letters

  JSIAM Letters 2 (0), 25-28, 2010

  The Japan Society for Industrial and Applied Mathematics

Keywords

• mixed precision
• singular value decomposition
• hyperplane constrained method

Details

• CRID: 1390282680278371456
• NII Article ID: 130000257636
• DOI: 10.14495/jsiaml.2.25
• ISSN: 18830617
• ISSN: 18830609
• Web Site: https://www.jstage.jst.go.jp/article/jsiaml/2/0/2_0_25/_pdf
• Text Lang: en
• Data Source:
  • JaLC
  • Crossref
  • CiNii Articles
  • KAKEN
• Abstract License Flag: Disallowed