Computing the polar decomposition and the related matrix sign function, has been a well-studied problem in numerical analysis for decades. More recently, it has emerged as an important subroutine in deep learning, particularly within the Muon optimization framework. However, the requirements in this setting differ significantly from those of traditional numerical analysis. In deep learning, methods must be highly efficient and GPU-compatible, but high accuracy is often unnecessary. As a resul...| papers.cool
在上文《msign算子的Newton-Schulz迭代（上）》中，我们试图为$\mathop{\text{msign}}$算子寻找更好的Newton-Schulz迭代，以期在有限迭代步数内能达到...| kexue.fm
最近在阅读时，遇到了一个关于最优多项式逼近的“等值振荡定理（Equioscillation Theorem）”，证明过程还涉及到无穷范数求导，感觉结论和证明都颇为新奇，特来记录一番。参考资料：《...| kexue.fm
在之前的《Muon优化器赏析：从向量到矩阵的本质跨越》、《Muon续集：为什么我们选择尝试Muon？》等文章中，我们介绍了一个极具潜力、有望替代Adam的新兴优化器——“Muon”。随着相关研究...| kexue.fm
事实上，除了写博客内容，在这几年里，笔者是花了相当一部分时间来做科学空间的“表面功夫”，为此还专门学了一点php、css和js。虽然不敢说精益求精，但总体来说网站的浏览体验应该比前几年要好得多。...| kexue.fm