Perfect density models cannot guarantee anomaly detection [article]

Charline Le Lan, Laurent Dinh
2021 arXiv   pre-print
Thanks to the tractability of their likelihood, some 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 and show that these quantities carry less meaningful information than
more » ... reviously thought, beyond estimation issues or the curse of dimensionality. We conclude that the use of these likelihoods for out-of-distribution detection relies on strong and implicit hypotheses, and highlight the necessity of explicitly formulating these assumptions for reliable anomaly detection.
arXiv:2012.03808v2 fatcat:kchdoxof6bhbxm6k3cnbopjuum