Abstract: Current state-of-the-art unsupervised anomalous sound detection (ASD) methods typically rely on manually annotated attribute information as labels, employing auxiliary classification tasks ...