English(EN) Convergence, design and training of continuous-time dropout as a random batch method

新方法近似连续时间Dropout在神经ODE中的应用

作者 PulseAugur 编辑部 · [1 个来源] · 2026-09-15 04:00

研究人员开发了一种新的随机批处理近似方法，用于控制微分方程中的连续时间Dropout。该技术为加性向量场提供了无偏近似，并在轨迹和分布层面都具有已证明的误差界限。该方法专为监督训练设计，为目标波动和最优值一致性提供了理论保证，数值实验证明了其在神经常微分方程上的有效性。 AI

影响引入了一种新颖的近似技术，用于训练神经ODE，可能提高效率和稳定性。

排序理由该集群包含一篇研究论文，详细介绍了连续时间Dropout在微分方程中的新方法。[lever_c_demoted from research: ic=1 ai=1.0]

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: 15 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群包含一篇研究论文，详细介绍了连续时间Dropout在微分方程中的新方法。[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: 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。

arXiv cs.LG TIER_1 English(EN) · Antonio \'Alvarez-L\'opez, Mart\'in Hern\'andez · 2026-09-15 04:00

连续时间Dropout作为随机批处理方法的收敛、设计和训练

arXiv:2510.13134v2 Announce Type: replace Abstract: We study continuous-time dropout in controlled differential equations. We introduce a random-batch approximation of additive vector fields. On each time interval of length $h$, a random subset of components is activated and resc…