测试测试:类分割异常检测中的分数方向不稳定性
Testing the Test: Score-Direction Instability in Class-Split Anomaly Detection
AI总结:
本文通过提出邻域类泄漏诊断方法,揭示了类分割异常检测协议在异常类与正常混合重叠时分数方向不稳定的问题,并建议将其视为几何依赖的压力测试。
摘要:
数据集内类分割评估被广泛用作完全无条件分布外异常检测的代理。我们表明,当留出的异常类在表示空间中与正常混合重叠时,该协议可能变得不适定。在这种情况下,异常分数可能趋近于随机甚至反转,且偏好的分数方向可能取决于未知的异常类。我们引入了一种简单的无训练诊断方法——邻域类泄漏,并表明它在Fashion-MNIST、CIFAR-10和Imagenette上,无论是在像素空间还是VAE潜在空间中,都能预测分数方向的不稳定性。我们的结果表明,类分割异常检测基准应被视为几何依赖的压力测试,而非异常检测能力的无条件证据。
英文摘要:
Within-dataset class-split evaluation is widely used as a proxy for fully unconditional out-of-distribution anomaly detection. We show that this protocol can become ill-posed when the held-out anomaly class overlaps the normal mixture in representation space. In this regime, anomaly scores may collapse toward chance or even invert, and the preferred score direction can depend on the unknown anomaly class. We introduce a simple training-free diagnostic, neighborhood class leakage, and show that it predicts score-direction instability across Fashion-MNIST, CIFAR-10, and Imagenette, in both pixel and VAE latent spaces. Our results suggest that class-split AD benchmarks should be treated as geometry-dependent stress tests rather than unconditional evidence of anomaly-detection ability.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。