English

Stability Improvements for Fast Matrix Multiplication

Numerical Analysis 2023-10-05 v1 Numerical Analysis

Abstract

We implement an Augmented Lagrangian method to minimize a constrained least-squares cost function designed to find polyadic decompositions of the matrix multiplication tensor. We use this method to obtain new discrete decompositions and parameter families of decompositions. Using these parametrizations, faster and more stable matrix multiplication algorithms can be discovered.

Keywords

Cite

@article{arxiv.2310.02794,
  title  = {Stability Improvements for Fast Matrix Multiplication},
  author = {Charlotte Vermeylen and Marc Van Barel},
  journal= {arXiv preprint arXiv:2310.02794},
  year   = {2023}
}
R2 v1 2026-06-28T12:40:24.327Z