English(EN) Muon-C: Operator-Aligned Muon for Convolutional Kernels

新的 Muon-C 优化器增强了卷积核性能

作者 PulseAugur 编辑部 · [1 个来源] · 2026-09-10 04:00

研究人员开发了 Muon-C，一种用于卷积核的新型优化器，它改进了现有的 Adam 和 unfolded Muon 等方法。通过将核动量表示为逐频率的通道传输矩阵，并独立极化这些块，Muon-C 确保更新保留在原始有限核支持内。这种新颖的几何结构带来了卓越的性能，在 CIFAR-10 流匹配上以更少的 FLOPs 实现了 9.87 FID，优于其前代产品，并在相同的调优预算下实现了 3.42 FID。 AI

影响引入了一种更有效的卷积神经网络优化方法，可能导致图像相关任务中更快的训练和更好的性能。

排序理由这是一篇详细介绍机器学习模型新优化技术的 ist 论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv stat.ML 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 46 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 这是一篇详细介绍机器学习模型新优化技术的 ist 论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, model release; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: Breaking (< 6h); Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv stat.ML TIER_1 English(EN) · Jiaxin Qing, Lexin Li · 2026-09-10 04:00

Muon-C：面向卷积核的算子对齐Muon

arXiv:2609.09676v1 Announce Type: cross Abstract: Muon replaces matrix momentum with an approximately orthogonal polar direction, but its geometry depends on the matrix representation. For convolution, standard unfolding describes a local patch map rather than the convolution ope…

本文如何被排名

报道来源 [1]

Muon-C：面向卷积核的算子对齐Muon

相关实体

相关话题