Revisiting “Rapid multiplication of rectangular matrices”

Roy Nissim*, Oded Schwartz

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

Abstract

What is the largest constant α such that for all ɛ > 0, an n × n and (Formula presented.) matrices can be multiplied in (Formula presented.) ? Coppersmith (1982) was the first to present a lower bound on α, showing that (Formula presented.). Coppersmith further proposes a second proof that leads to the bound (Formula presented.), and combined with a newer construction by Coppersmith and Winograd (1982), the second bound improves to (Formula presented.). We revisit this work, show that the second proof is incorrect, and propose an alternative one. Our alternative proof uses a recursive construction, based on the Schönhage’s multiplication stacking technique, that converges to the same bound on α.

Original languageEnglish
Article number2388334
JournalResearch in Mathematics
Volume11
Issue number1
DOIs
StatePublished - 2024

Bibliographical note

Publisher Copyright:
© 2024 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group.

Keywords

  • algebraic complexity
  • Fast matrix multiplication
  • lower bounds
  • rectangular matrices
  • tensor rank

Fingerprint

Dive into the research topics of 'Revisiting “Rapid multiplication of rectangular matrices”'. Together they form a unique fingerprint.

Cite this