English(EN) Heteroscedasticity of Denoising Score Matching with Generalised Smooth Noise

新研究揭示去噪分数匹配中固有的不稳定性

作者 PulseAugur 编辑部 · [1 个来源] · 2026-08-05 04:00

Juyan Zhang及其同事在arXiv上发表的一篇新论文探讨了去噪分数匹配（DSM）中固有的异方差性。DSM是一种广泛用于生成模型的技巧。研究人员证明，DSM的目标函数会导致模型参数方差不可预测，具体取决于噪声水平和数据几何。他们提出了一种理想的加权函数来均衡这种方差，从而创建DSM的同方差泛化，并推导了一种实用的近似方法用于训练，以减少梯度方差。 AI

影响这项研究为改进扩散模型和其他生成式AI系统的稳定性和性能提供了理论基础。

排序理由该集群包含一篇详细介绍机器学习理论发现和提出方法的学术论文。[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: 0 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群包含一篇详细介绍机器学习理论发现和提出方法的学术论文。[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: 79 days old; Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv stat.ML TIER_1 English(EN) · Juyan Zhang, Rhys Newbury, Xinyang Zhang, Tin Tran, Dana Kulic, Michael Burke · 2026-08-05 04:00

具有广义平滑噪声的去噪分数匹配的异方差性

arXiv:2508.01597v2 Announce Type: replace-cross Abstract: Score Matching (SM) is a powerful framework for estimating the log-density derivatives of a distribution without calculating its normalizing constants. This capability has made it a cornerstone across multiple domains, fro…

本文如何被排名

报道来源 [1]

具有广义平滑噪声的去噪分数匹配的异方差性

相关实体

相关话题