Skip to main navigation Skip to search Skip to main content

UNIFIED SMW-LIKE IDENTITIES OF LOW-RANK UPDATES FOR GENERALIZED INVERSES AND PSEUDOINVERSES

Research output: Contribution to journalArticlepeer-review

Abstract

We present new low-rank-update identities for generalized inverses and pseudoinverses of rectangular matrices, unifying previous generalizations of the Sherman–Morrison–Woodbury (SMW) identity and Riedel's rank-augmentation formulas. First, we establish generalized SMW identities for \{1\}-inverses and pseudoinverses under less restrictive conditions when the matrix rank is preserved. Second, we further generalize Riedel's formulas for rank augmentation to pseudoinverses of rectangular matrices and to specific classes of generalized inverses (namely, those involving \{1,3\}- and \{1,4\}-inverses) when matrix ranges are altered by the update. Finally, we introduce unified low-rank-update identities that encompass both cases. These identities retain SMW's efficiency, extend its applicability to rectangular matrices, and offer new tools for updating generalized inverses in various applications.

Original languageEnglish
Pages (from-to)1917-1929
Number of pages13
JournalSIAM Journal on Matrix Analysis and Applications
Volume46
Issue number3
DOIs
StatePublished - 2025

Keywords

  • generalized inverse
  • low-rank update
  • Moore–Penrose pseudoinverse
  • rank augmentation
  • rank-deficient least squares
  • Sherman–Morrison–Woodbury identity

Fingerprint

Dive into the research topics of 'UNIFIED SMW-LIKE IDENTITIES OF LOW-RANK UPDATES FOR GENERALIZED INVERSES AND PSEUDOINVERSES'. Together they form a unique fingerprint.

Cite this