The Signal in the Noise: OOD Detection Through Goodness-of-Fit Testing in Factorised Latent Spaces
For practitioners using deep generative models, this provides a principled and computationally efficient OOD detection method that overcomes the known failure of likelihoods.
The paper addresses the unreliability of likelihood-based OOD detection in deep generative models by proposing SITN, a method that uses goodness-of-fit testing in the latent space of normalising flows. SITN achieves strong OOD detection performance with strict false positive rate control and no need for OOD data, outperforming likelihood-based methods on standard benchmarks.
Deep generative models offer a natural foundation for out-of-distribution (OOD) detection, yet prior work has shown that their assigned likelihoods are notoriously unreliable indicators for in- vs out-of-distribution data. In this paper, we address this problem by leveraging the diffeomorphic and mass-preserving properties of continuous normalising flows. Our analysis shows that OOD samples are mapped to noise samples that are highly atypical under the noise prior in ways not captured by the likelihood. Based on this observation, we propose a new method -- Signal in the Noise (SITN) -- for OOD detection on the single-sample level. SITN requires no access to OOD data, incurs minimal computational overhead, and provides strict control of false positive rates. Comprehensive evaluations through standard benchmarks and synthetic perturbations highlight the method's effectiveness and the absence of the complexity bias inherent to likelihood-based methods.