arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2609.32505math.NAcs.LGcs.NA

通过极分解在GPU上实现快速可微SVD

Fast Differentiable SVD on GPU via Polar Decomposition

  • HSE University(高等经济大学)

机构由 AI 辅助整理,请以论文原文为准。

Uliana Parkina, Askar Tsyganov, Sergei Kudriashov, Sergey Samsonov, Maxim Rakhuba

AI总结:

本文提出基于极分解的GPU SVD流水线,利用Newton-Schulz迭代实现2倍加速,并推导数值稳定的反向传播以获得完全可微的SVD,开源于PyTorch和JAX。

AI中文摘要:

我们提出了一种完全面向GPU的SVD(奇异值分解)流水线,其基于极分解,并受到仅依赖矩阵乘法的迭代方法(如Newton-Schulz迭代)的启发。我们证明,与标准实现相比,该方法可实现高达2倍的加速。此外,我们推导了极分解的数值稳定的反向传播,并利用它获得完全可微的SVD。我们的方法以PyTorch和JAX两种框架的开源实现形式发布:此https URL。

英文摘要:

We present a fully GPU-oriented SVD pipeline based on polar decomposition, motivated by iterative methods that rely solely on matrix multiplications, such as the Newton-Schulz iteration. We show that this approach enables up to a $2\times$ speedup compared to standard implementations. Furthermore, we derive a numerically stable backward pass for the polar decomposition and leverage it to obtain a fully differentiable SVD. Our methods are released as open-source implementations in both PyTorch and JAX: https://github.com/fallnlove/cans_svd.

↑