阅读：General Quantification of Covariate and Concept Shifts
- 模式: 沉浸阅读高引用论文 (roll=35)
- 时间: 2026-09-11 19:09:20
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
Lyco 推眼镜，翻开草稿纸，清了清嗓子：
---
1. 背景：理论与实践的“两张皮”，老定义还“碎”了
同学们，咱们先搞清背景。分布偏移是机器学习的“心头大患”，训练集分布 $P$ 和测试集分布 $Q$ 对不上，模型就容易翻车。虽然理论界早有“泛化误差界”这把尺子，但既不够通用（只盯着协变量偏移、标签偏移这些理想化场景），又算不出来（理论量依赖真实分布密度比，样本里根本估不准）。更尴尬的是，论文一上来就给传统“概念偏移”定义来了个“致命一击”：一旦源域和目标域的支撑集不匹配（比如训练集里没见过的样本跑测试集来了），旧定义直接失效/发散，根本没法用。这就像量身高的尺子，人一站上去就断了，还怎么量？
---
2. 核心招数：引入“熵正则最优传输”，重造“$\gamma^*$-概念偏移”
既然旧尺子不好使，作者祭出大杀器——熵正则最优传输。别怕，初中生也能懂的直觉：在源域和目标域之间搭座“软桥”，允许质量分散、允许噪声、允许支撑集不重叠。基于这个桥，他们定义了$\gamma^$-概念偏移——用最优传输计划 $\gamma^$ 下的条件分布差异来度量“概念变了多少”。
妙就妙在：这玩意儿天然统一了协变量偏移和概念偏移，推导出的误差界适用广义损失函数、任意标签空间、甚至随机标签。不管你是分类、回归还是结构化预测，这把尺子全能量，而且理论上严格成立，不再怕支撑集不匹配。
---