English(EN) Generative SDEs with Trainable Diffusion and Density-Based Losses

新方法使用基于密度的损失优化神经SDE

作者 PulseAugur 编辑部 · [1 个来源] · 2026-11-02 05:00

研究人员开发了一种新的方法来优化神经随机微分方程（SDE），通过联合训练漂移和扩散场以及初始高斯分布。该方法利用直接考虑终端密度的损失。在二维和八维设置中的实验表明，该方法学习了与时间无关的场，并且与具有相似参数数量的基于分数的生成器相比，产生的离模样本显著减少。漂移分量将样本引导至模态，而扩散分量则有助于在它们之间分配概率。 AI

影响这项研究可能通过改进SDE的优化来带来更高效、更准确的生成模型。

排序理由该集群包含一篇详细介绍优化神经SDE新方法的论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.LG 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: 1 / 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优化神经SDE新方法的论文。[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, other; 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: 1 days old; Coverage has settled into its steady-state source set.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.LG TIER_1 English(EN) · Angxiu Ni · 2026-11-02 05:00

Generative SDEs with Trainable Diffusion and Density-Based Losses

arXiv:2610.41243v1 Announce Type: new Abstract: We develop a divergence-kernel method for jointly optimizing neural SDE drift and diffusion fields and the initial Gaussian distribution, using losses that depend directly on the terminal density. In two- and eight-dimensional Gauss…

本文如何被排名

报道来源 [1]

Generative SDEs with Trainable Diffusion and Density-Based Losses

相关实体

相关话题