Input complexity and out-of-distribution detection with likelihood-based generative models
This work addresses robustness issues in machine learning systems by improving OOD detection, though it appears incremental as it builds on existing likelihood-based methods with a complexity adjustment.
The paper tackled the problem of out-of-distribution (OOD) detection using likelihood-based generative models by addressing the excessive influence of input complexity on likelihoods, resulting in a parameter-free OOD score that performs comparably or better than existing approaches across various datasets and models.
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 significantly 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 comparison. 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.