Abstract: Anomaly detection (AD) is typically regarded as an unsupervised learning task, where the training data either do not contain any anomalous samples or contain only a few unlabeled anomalous ...