Preprints, Working Papers, ... Year : 2025

Learning Permutations in Monarch Factorization

Abstract

In order to reduce the quadratic cost of matrixvector multiplications in dense and attention layers, Monarch matrices have been recently introduced, achieving a sub-quadratic complexity. It consists in factorizing a matrix using fixed permutations and learned block diagonal matrices, at the price of a small performance drop. We propose a more general model where some permutations are learned. The optimization algorithm explores the space of permutations using a Straight-Through Estimator (STE) inspired by the support exploration algorithm designed for sparse support recovery. Our experimental results demonstrate performance improvement in the context of sparse matrix factorization and of end-to-end sparse learning.
Fichier principal
Vignette du fichier
ICASSP_2025_Monarch.pdf (317) Télécharger le fichier
Origin Files produced by the author(s)

Dates and versions

hal-04887483 , version 1 (17-01-2025)

Identifiers

  • HAL Id : hal-04887483 , version 1

Cite

Mimoun Mohamed, Valentin Emiya, Caroline Chaux. Learning Permutations in Monarch Factorization. 2025. ⟨hal-04887483⟩
0 View
0 Download

Share

More