一种有原则的无监督异常检测方法
A Principled Approach to Unsupervised Anomaly Detection
- Imperial College London（帝国理工学院）
- Friedrich-Alexander University Erlangen-Nürnberg（弗里德里希-亚历山大大学埃尔朗根-纽伦堡）
- Nanyang Technological University（南洋理工大学）
机构由 AI 辅助整理，请以论文原文为准。
AI总结：
提出将无监督异常检测重构为贝叶斯逆问题，以能量分数作为异常分数，统一现有方法并改进MVTec AD数据集上的AUROC，同时提供病理估计。
AI中文摘要：
传统的无监督异常检测（UAD）方法旨在标记或定位偏离规范分布的偏差，而忽略了异常背后的生成机制。然而，异常的性质往往与其存在同样重要。我们将UAD重新表述为一个贝叶斯逆问题，其目标是推断每个观测值最可能的损坏原因。我们的框架产生了一个概率异常分数，作为推断损坏参数的能量，并作为开发新UAD算法的有原则的配方。我们将几种现有方法推导为一般框架的实例，每种方法对应于不同建模选择下的相同能量分数。在实验上，我们在受控环境中研究了框架的组成部分，并通过调整底层损坏模型，在MVTec AD数据集上将对象类别AUROC提高了2.3%。最后，我们在脑MRI基准上验证了该框架，实现了强大的检测性能，同时产生了病理强度、偏置和几何形状的估计。代码可在https URL获取。
英文摘要：
Traditional unsupervised anomaly detection (UAD) methods are designed to flag or localise deviations from a normative distribution, ignoring the underlying generative mechanisms of the anomalies. Yet the nature of an anomaly is often as important as its presence. We reformulate UAD as a Bayesian inverse problem, in which the objective is to infer the most probable corruption responsible for each observation. Our framework yields a probabilistic anomaly score as the energy of the inferred corruption parameters, and serves as a principled recipe for developing new UAD algorithms. We derive several existing methods as instances of the general framework, each corresponding to the same energy score under different modelling choices. Experimentally, we study the framework's components in a controlled setting, and improve object-class AUROC on the MVTec AD dataset by 2.3% by adapting the underlying corruption model. Finally, we validate the framework on a brain MRI benchmark, achieving strong detection performance while producing estimates of pathology intensity, bias, and geometry. Code is available at https://github.com/jgmyles/inverse-uad.
