用核锚定局部性正则化缓解固定目标异常检测器的收敛坍缩

View PDF HTML (experimental)

Abstract:A family of tabular anomaly detectors trains a neural map toward a fixed target under squared-error loss and scores anomalies by the test-time residual; contraction matching, one-step rectified flow, and reconstruction autoencoders all fit this template. We characterize a convergence collapse: better optimization makes the detector worse. At convergence, the learned map tracks the target even off-distribution, so the residual signal vanishes on anomalies as well as on normal data. These detectors therefore rely on implicit non-convergence (early stopping, capacity caps) to retain signal. We argue this is structural: effective anomaly detection requires a locality constraint that blocks unconstrained extrapolation. Classical detectors (kNN, KDE, isolation forests, LOF) enforce locality explicitly; fixed-target neural detectors do not. We formalize the connection by showing that the kernel-regression analog of a fixed-target detector is a finite-bandwidth Nadaraya-Watson smoother, which we call Kernel Contraction Matching (KCM). KCM is closed-form, training-free, and CPU-efficient, yet matches established neural baselines on ADBench. Building on this bridge, we introduce the Kernel-Anchored Regularizer (KAR), which penalizes deviation of the neural prediction from a kernel-weighted average of training targets. Across collapse-prone ADBench datasets and three backbones, KAR mitigates collapse and improves AUROC under prolonged training.

Comments:	Accepted at CIKM 2026 (oral). 11 pages
Subjects:	Machine Learning (cs.LG)
Cite as:	arXiv:2610.02345 [cs.LG]
	(or arXiv:2610.02345v1 [cs.LG] for this version)
	https://doi.org/10.48550/arXiv.2610.02345 arXiv-issued DOI via DataCite (pending registration)

Submission history

From: José Costa [view email]
[v1] Thu, 1 Oct 2026 18:16:58 UTC (893 KB)