Paper Overview
- Research area: Machine Learning
- Author: Stephen Becker
- Published: 2026-07-24
- arXiv: 2607.22484
- Singular value soft-thresholding is reduced to computing the matrix polar decomposition.
- The polar decomposition can be computed with GPU-friendly algorithms, unlike direct SVD-based methods.
- Empirical results show a significant GPU speed-up over the standard SVD approach.
- Robustness analysis is deferred to future work; due to the discontinuity of the sign function, the method is likely best suited to low-accuracy applications.
Abstract (translated)
Singular value soft-thresholding can be computed via a reduction to the matrix polar decomposition, which allows one to exploit GPU-friendly algorithms for computing the polar decomposition. Empirically, there is a significant speed-up on GPUs compared to the standard approach using the SVD. We leave the investigation of robustness to future work, but note that due to the discontinuous nature of the sign function, the reduction to the polar decomposition is likely only suitable for low-accuracy applications.
Key Points
*Auto-collected on 2026-07-28.*