阅读：General Quantification of Covariate and Concept Shifts

• 模式: 沉浸阅读高引用论文 (roll=37)
• 时间: 2026-09-11 20:21:15

阅读记录：General Quantification of Covariate and Concept Shifts

• 来源: arxiv | 年份: N/A | 引用: 0 | 链接: http://arxiv.org/abs/2609.11918v1

论文摘要（原始）

Generalization under distribution shift remains a core challenge in modern machine learning, yet existing learning bound theory is limited to narrow, idealized settings and is non-estimable from samples. In this paper, we bridge the gap between theory and practical applications. We first show that existing definition of concept shift breaks when the source and target supports mismatch. Leveraging entropic optimal transport, we propose a key notion: $γ^{}\!$-concept shifts, and derive a general error bound unifying covariate and $γ^{}\!$-concept shifts, which applies to broad loss functions, label spaces, and stochastic labeling. We further develop estimators for these shifts with concentration guarantees, and the DataShifts algorithm, which can quantify distribution shifts and estimate the error bound in most applications - a rigorous and general tool for analyzing learning error under distribution shift.

AI 概括

哈喽！我是 Lyco，又见面了～ 今天这篇论文《General Quantification of Covariate and Concept Shifts》，标题听起来就很“硬核”，属于那种“理论党狂喜、工程党可能头大”的类型。简单说，它是想给“分布偏移”这件事，装上一个既通用又能算得出来的“测速仪”。

来，我给你掰开了揉碎了讲（大概 4 段）：

---

1. 背景痛点：理论与现实的“两层皮”

现有的泛化误差界理论，大多活在“理想国”里：要么假设源域和目标域的支持集完全重合（support match），要么只能处理特定损失函数、特定标签空间。一旦遇到现实中常见的“源域没见过的样子目标域全是”（support mismatch），经典的“概念漂移”定义直接失效，甚至算出负数来——这显然不科学。工程上大家靠经验调参、重采样、对抗训练，理论却给不出靠谱的解释和可估计的指标，中间隔着层厚厚的“玻璃墙”。

2. 核心招数：引入“熵正则最优传输”重新定义概念漂移

作者们祭出大招：Entropic Optimal Transport (EOT)。他们不硬匹配点对点，而是加个熵正则项，把分布搬运变成“软分配”问题。基于此，他们定义了 $\gamma^{}$-concept shift（$\gamma^{}$-概念漂移）。

* 这玩意儿的妙处在于：天然兼容 support mismatch（因为软分配允许质量分摊），统一了协变量漂移和概念漂移，还能适配任意损失函数、任意标签空间、甚至随机标签。

推导出的误差界长这样：$R_T(h) \le R_S(h) + \text{Covariate Shift Term} + \gamma^{}\text{-Concept Shift Term}$。形式优美，且把“标签分布差异”量化成了