基本信息
- 来源: arxiv
- 原始来源: http://arxiv.org/abs/2609.31612v1
- 发布域名: arxiv.org
- 分类: stat.ML
- 作者: Zhifeng Chen、Chenyang Jiang、Yazhen Wang
要点解读
这是什么
这篇论文把优化中的一阶平稳性概念迁移到扩散模型，给出基于随机微分方程的逆向扩散在强凸势条件下相对于 Fisher 散度的收缩性，并给出离散化后平均一阶平稳性界的理论结果。
用在哪里
适用于研究扩散模型收敛性、设计新采样算法或进行理论分析的研究者，尤其在需要评估非凸分布采样质量时。
可以推断的
推测：该理论不依赖特定模型结构，只要满足所给的强凸势条件即可使用。
推测：在实际离散采样时，平均梯度范数的上界可作为收敛性的参考指标。
来源摘要/节选
Recent literature has shown a strong connection between optimization and sampling. We develop the corresponding first-order theory for diffusion models. First, the SDE-based reverse-time flows of overdamped and underdamped Langevin diffusions contract relative Fisher divergences at explicit exponential rates whenever the stationary potential of the forward process is strongly convex—a condition on the noising process one chooses, not on the data. This is a unique advantage of SDE-based reverse diffusion, absent in the reverse process based on ODEs. Second, we incorporate discretization and establish averaged first-order stationarity bounds—the sampling analog of averaged gradient-norm guarantees in nonconvex optimization—for samplers of both overdamped and underdamped diffusion models. As in nonconvex optimization, the convexity-free certificate is local: it guarantees score consistency, not global mode weights.
来源说明
当前保存的是来源摘要，不代表论文全文。请以原始来源为准。
「要点解读」由 AI Stack 依据上方已保存内容整理，不代表来源的完整表述；标注「推测：」的判断来自编辑，不是来源陈述。