arXiv：cs.LG（机器学习，全量分类）· Gia Huy Thai, Nguyen Thai Anh·· 14 小时前AI 评分33
少样本异常检测阈值需要多少类别？无分布认证的可行性边界
How Many Categories Are Enough? Distribution-Free Certification Limits for Few-Shot Anomaly Thresholds
AI 导读
研究用冻结的 DINOv2 PCA 残差排序器在 15 个 MVTec 和 12 个 VisA 类别上验证，仅靠目标类别图像的 LOIO 校准受分辨率限制且对分布偏移脆弱，在 α=0.20、k=4 时高斯损坏 MVTec 上的实测 FAR 达 0.341，是标称水平的 1.7 倍。
正文
Abstract:Few-shot anomaly detectors are judged by ranking metrics, yet deployment requires an alarm threshold with a controlled false-alarm rate (FAR). We ask how much normal evidence, in images or category units, is needed to certify such a threshold for an unseen category. Using a frozen DINOv2 principal component analysis (PCA) residual ranker on 15 MVTec and 12 VisA categories under four corruption types, we show that target-only leave-one-image-out (LOIO) calibration is resolution-limited and shift-fragile: rank values cannot fall below $1/(k+1)$, and at the attainable level $\alpha=0.20$, empirical FAR reaches 0.341 on Gaussian-corrupted MVTec at $k=4$, 1.7 times the nominal level. A category-count feasibility calculus is then derived: even with all-zero category losses and no multiplicity charged, any deterministic, uniformly valid, distribution-free 95% upper confidence bound (UCB) requires at least 14, 29, and 59 independent and identically distributed (iid) category draws at $\alpha=0.20$, $0.10$, and $0.05$; these counts are necessary but not sufficient. The Cross-category Reliability Estimation with Source Support (CRESS) protocol splits source categories into disjoint reference, proposal, and certification roles. With only three or four certification categories, all 960 frozen configurations return the fail-closed threshold $\tau^\star=0$, and the smallest category-level UCB is 0.950. Image-unit analyses of the same archive select positive thresholds in 36.7% to 60.3% of target cells; these bounds hold for the selected source mixture, not for the marginal risk of a new-category draw. The contribution is a quantitative feasibility boundary and an estimand-aware protocol specifying when source evidence can, and cannot, support a transferable reliability claim.
| Subjects: | Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.00236 [cs.LG] |
| (or arXiv:2610.00236v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.00236 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Nguyen Thai Anh [view email]
[v1]
Wed, 23 Sep 2026 03:24:09 UTC (1,593 KB)
来源：arXiv：cs.LG（机器学习，全量分类） · arxiv.org