Conference paper Open Access
Serra, Joan; Álvarez, David; Gómez, Vicenç; Slizovskaia , Olga; F. Núñez, José; Luque, Jordi
Likelihood-based generative models are a promising resource to detect out-of- distribution (OOD) inputs which could compromise the robustness or reliability of a machine learning system. However, likelihoods derived from such models have been shown to be problematic for detecting certain types of inputs that sig- nificantly differ from training data. In this paper, we pose that this problem is due to the excessive influence that input complexity has in generative models’ likelihoods. We report a set of experiments supporting this hypothesis, and use an estimate of input complexity to derive an efficient and parameter-free OOD score, which can be seen as a likelihood-ratio, akin to Bayesian model compari- son. We find such score to perform comparably to, or even better than, existing OOD detection approaches under a wide range of data sets, models, model sizes, and complexity estimates.