Why the Counterintuitive Phenomenon of Likelihood Rarely Appears in Tabular Anomaly Detection with Deep Generative Models?
This addresses a key reliability concern for anomaly detection in tabular data, providing evidence that deep generative models are less prone to counterintuitive likelihood assignments compared to image domains, which is incremental but clarifies a domain-specific issue.
The paper tackles the problem of whether deep generative models assign higher likelihoods to anomalous data in tabular anomaly detection, finding that such counterintuitive behavior is consistently rare across 47 tabular datasets and 10 embedding datasets, with likelihood-only detection using normalizing flows proving practical and reliable.
Deep generative models with tractable and analytically computable likelihoods, exemplified by normalizing flows, offer an effective basis for anomaly detection through likelihood-based scoring. We demonstrate that, unlike in the image domain where deep generative models frequently assign higher likelihoods to anomalous data, such counterintuitive behavior occurs far less often in tabular settings. We first introduce a domain-agnostic formulation that enables consistent detection and evaluation of the counterintuitive phenomenon, addressing the absence of precise definition. Through extensive experiments on 47 tabular datasets and 10 CV/NLP embedding datasets in ADBench, benchmarked against 13 baseline models, we demonstrate that the phenomenon, as defined, is consistently rare in general tabular data. We further investigate this phenomenon from both theoretical and empirical perspectives, focusing on the roles of data dimensionality and difference in feature correlation. Our results suggest that likelihood-only detection with normalizing flows offers a practical and reliable approach for anomaly detection in tabular domains.