PRISM: Distribution-free Adaptive Computation of Matrix Functions for Accelerating Neural Network Training

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Yang, Shenghao, Wang, Zhichao, Balabanov, Oleg, Erichson, N. Benjamin, Mahoney, Michael W.
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912860136800256
author Yang, Shenghao
Wang, Zhichao
Balabanov, Oleg
Erichson, N. Benjamin
Mahoney, Michael W.
author_facet Yang, Shenghao
Wang, Zhichao
Balabanov, Oleg
Erichson, N. Benjamin
Mahoney, Michael W.
contents Matrix functions such as square root, inverse roots, and orthogonalization play a central role in preconditioned gradient methods for neural network training. This has motivated the development of iterative algorithms that avoid explicit eigendecompositions and rely primarily on matrix multiplications, making them well suited for modern GPU accelerators. We present PRISM (Polynomial-fitting and Randomized Iterative Sketching for Matrix functions computation), a general framework for accelerating iterative algorithms for computing matrix functions. PRISM combines adaptive polynomial approximation with randomized sketching: at each iteration, it fits a polynomial surrogate to the current spectrum via a sketched least-squares problem, adapting to the instance at hand with minimal overhead. We apply PRISM to accelerate Newton-Schulz-like iterations for matrix square roots and orthogonalization, which are core primitives in machine learning. Unlike prior methods, PRISM requires no explicit spectral bounds or singular value estimates; and it adapts automatically to the evolving spectrum. Empirically, PRISM accelerates training when integrated into Shampoo and Muon optimizers.
format Preprint
id arxiv_https___arxiv_org_abs_2601_22137
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle PRISM: Distribution-free Adaptive Computation of Matrix Functions for Accelerating Neural Network Training
Yang, Shenghao
Wang, Zhichao
Balabanov, Oleg
Erichson, N. Benjamin
Mahoney, Michael W.
Machine Learning
Artificial Intelligence
Numerical Analysis
Optimization and Control
Matrix functions such as square root, inverse roots, and orthogonalization play a central role in preconditioned gradient methods for neural network training. This has motivated the development of iterative algorithms that avoid explicit eigendecompositions and rely primarily on matrix multiplications, making them well suited for modern GPU accelerators. We present PRISM (Polynomial-fitting and Randomized Iterative Sketching for Matrix functions computation), a general framework for accelerating iterative algorithms for computing matrix functions. PRISM combines adaptive polynomial approximation with randomized sketching: at each iteration, it fits a polynomial surrogate to the current spectrum via a sketched least-squares problem, adapting to the instance at hand with minimal overhead. We apply PRISM to accelerate Newton-Schulz-like iterations for matrix square roots and orthogonalization, which are core primitives in machine learning. Unlike prior methods, PRISM requires no explicit spectral bounds or singular value estimates; and it adapts automatically to the evolving spectrum. Empirically, PRISM accelerates training when integrated into Shampoo and Muon optimizers.
title PRISM: Distribution-free Adaptive Computation of Matrix Functions for Accelerating Neural Network Training
topic Machine Learning
Artificial Intelligence
Numerical Analysis
Optimization and Control
url https://arxiv.org/abs/2601.22137