English(EN) On the asymptotic behaviour of $\alpha$-divergence generalized posteriors

研究论文分析$\alpha$-散度广义后验

作者 PulseAugur 编辑部 · [1 个来源] · 2026-10-27 04:00

一篇新研究论文探讨了当使用$\alpha$-散度进行正则化时，广义变分推断解的渐进行为。研究表明，在某些条件下，该问题的最小化值可以被明确确定并表现出重标度性质。论文详细阐述了该重标度最小化值的极限如何取决于$\alpha$的值，从而导致不同的分布，如Student-t或Pearson II型，并影响后验宽度相对于采样误差。Kullback–Leibler散度（对应于$\alpha=1$）被确定为后验宽度与采样尺度匹配且极限为高斯分布的唯一情况。 AI

影响为可能影响未来AI模型开发的统计推断方法提供了理论见解。

排序理由学术论文发表在arXiv上，详细介绍了理论统计分析。[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 →

$研究论文分析$\alpha$-散度广义后验$

本文如何被排名

Signal score: 22 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 学术论文发表在arXiv上，详细介绍了理论统计分析。[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。

报道来源 [1]

arXiv stat.ML TIER_1 English(EN) · Ioar Casado-Telletxea, Omar Rivasplata · 2026-10-27 04:00

关于$\alpha$-散度广义后验分布的渐近行为

arXiv:2610.35330v1 Announce Type: cross Abstract: We study the asymptotic behaviour of the solutions to the unconstrained generalized variational inference problem regularized by an $\alpha$-divergence. Under mild conditions, the minimizer is explicit up to a scalar normalizing c…

本文如何被排名

报道来源 [1]

关于$\alpha$-散度广义后验分布的渐近行为

相关实体

相关话题