On MUON optimization: From non-convergence to an error analysis with Polar Express and the Newton-Schulz polynomial from implementations

📄 arXiv: 2608.04607v1 📥 PDF

作者: Thang Do, Steffen Dereich, Arnulf Jentzen

分类: math.OC, cs.LG

发布日期: 2026-08-05

备注: 82 pages


💡 一句话要点

提出MUON优化器的广义变体以解决收敛性问题

🎯 匹配领域: 支柱九:具身大模型 (Embodied Foundation Models)

关键词: 随机梯度下降 深度学习 优化算法 牛顿-舒尔茨 收敛性分析 广义MUON 人工智能 逻辑回归

📋 核心要点

  1. 现有的MUON优化器在处理特定随机优化问题时存在收敛性不足的挑战,尤其是在梯度步骤趋向无穷时。
  2. 本文提出了一种广义MUON优化器,允许使用任意数量的广义牛顿-舒尔茨步骤,增强了优化过程的灵活性和适应性。
  3. 通过对多个具体示例的分析,本文展示了广义MUON优化器在收敛速率和优化效果上的显著提升。

📝 摘要(中文)

随机梯度下降(SGD)优化方法是深度神经网络(DNN)训练的标准工具。然而,在许多人工智能系统中,尤其是大型语言模型(LLMs),通常采用加速的SGD变体。本文提出了一种MUON优化器的广义变体,涉及任意数量的广义牛顿-舒尔茨步骤,并提供了相应的误差分析,揭示了在特定随机优化问题中MUON的收敛性不足。我们还通过具体示例展示了该优化器的有效性和收敛速率。

🔬 方法详解

问题定义:本文旨在解决MUON优化器在特定随机优化问题中收敛性不足的问题,尤其是在梯度步骤数量趋向无穷时,MUON未能收敛到解的现象。

核心思路:提出了一种广义MUON优化器,允许使用任意数量的广义牛顿-舒尔茨步骤,以提高收敛性和优化效率。通过引入多项式的高阶形式,增强了优化器的灵活性。

技术框架:该优化器的整体架构包括多个阶段:首先定义优化问题,然后应用广义牛顿-舒尔茨步骤进行迭代,最后通过误差分析评估收敛性和性能。

关键创新:最重要的创新在于引入了广义牛顿-舒尔茨步骤,使得MUON优化器能够适应更复杂的优化场景,显著提升了收敛性和效率。

关键设计:在设计中,关键参数包括广义牛顿-舒尔茨步骤的数量和多项式的阶数,损失函数采用标准的随机优化损失形式,网络结构则依赖于具体的应用场景。

🖼️ 关键图片

img_0
img_1
img_2

📊 实验亮点

实验结果表明,广义MUON优化器在多个具体示例中表现出优越的收敛性和效率,尤其是在处理二元分类的$ ext{l}_2$正则化逻辑回归问题时,相较于传统SGD方法,收敛速率提升显著。

🎯 应用场景

该研究的潜在应用领域包括深度学习模型的训练,尤其是在大型语言模型和复杂神经网络的优化中。通过提高收敛性和效率,广义MUON优化器能够在实际应用中显著提升模型性能,推动人工智能技术的发展。

📄 摘要(原文)

Stochastic gradient descent (SGD) optimization methods are the standard instruments for the training of deep neural networks (DNNs). In many relevant artificial intelligence (AI) systems - such as popular large language models (LLMs)-not the standard SGD scheme is used as the optimization method but instead suitable accelerated variants of SGD are employed. One of the most popular methods of such accelerated SGD variants is the momentum orthogonalized by Newton-Schulz (MUON) optimizer proposed by Jordan et al. in 2024. The MUON optimizer exploits the special matrix structure of the weight parameters in the training of the DNNs and, in its original form, employs five Newton-Schultz (NS) matrix steps in each MUON iteration. In this work we propose and study a generalized variant of the MUON optimizer involving an arbitrary number of generalized NS steps with polynomials of possibly arbitrary high degree. The considered optimizer covers MUON with the original NS polynomial as well as MUON combined with the recently proposed Polar Express method as special cases. For a simple class of stochastic optimization problems (SOPs) we show for almost every mini-batch size that MUON fails to converge to the solution of the SOP as the number of gradient steps converges to infinity. We also establish an error analysis for MUON with the generalized NS steps that provides convergence rates in terms of the number of gradient steps and in terms of the size of the mini-batch. We illustrate our general error analysis for MUON in the case of several concrete examples including quadratic stochastic optimization problems (SOPs) as well as $\ell_2$ regularized logistic regression for binary classification.