Perfect density models cannot guarantee anomaly detection
This work identifies a fundamental limitation in using likelihoods from generative models for anomaly detection, impacting researchers and practitioners relying on these methods.
This paper investigates the use of likelihood values from deep generative models for anomaly detection. It demonstrates that these likelihoods, even from perfect density models, carry less meaningful information for anomaly detection than previously assumed, due to issues beyond estimation or dimensionality.
Thanks to the tractability of their likelihood, several deep generative models show promise for seemingly straightforward but important applications like anomaly detection, uncertainty estimation, and active learning. However, the likelihood values empirically attributed to anomalies conflict with the expectations these proposed applications suggest. In this paper, we take a closer look at the behavior of distribution densities through the lens of reparametrization and show that these quantities carry less meaningful information than previously thought, beyond estimation issues or the curse of dimensionality. We conclude that the use of these likelihoods for anomaly detection relies on strong and implicit hypotheses, and highlight the necessity of explicitly formulating these assumptions for reliable anomaly detection.