Revisiting Out-of-Distribution Detection: A Simple Baseline is Surprisingly Effective

Julian Bitterwolf, Alexander Meinke, Maximilian Augustin, Matthias Hein

Keywords: out-of-distribution detection, robustness, OOD

Abstract: It is an important problem in trustworthy machine learning to recognize out-of-distribution (OOD) inputs which are inputs unrelated to the in-distribution task. Many out-of-distribution detection methods have been suggested in recent years. The goal of this paper is to recognize common objectives as well as to identify the implicit scoring functions of different OOD detection methods. In particular, we show that binary discrimination between in- and (different) out-distributions is equivalent to several different formulations of the OOD detection problem. When trained in a shared fashion with a standard classifier, this binary discriminator reaches an OOD detection performance similar to that of Outlier Exposure. Moreover, we show that the confidence loss which is used by Outlier Exposure has an implicit scoring function which differs in a non-trivial fashion from the theoretically optimal scoring function in the case where training and test out-distribution are the same, but is similar to the one used when training with an extra background class. In practice, when trained in exactly the same way, all these methods perform similarly and reach state-of-the-art OOD detection performance.

One-sentence Summary: We analyze different training methods and scoring functions for out-of-distribution detection and identify a binary discriminator as a critical implicit component in different OOD detection approaches.