Towards Maximizing the Representation Gap between In-Domain & Out-of-Distribution Examples
Jay Nandy, Wynne Hsu, Mong Li Lee
Abstract
Among existing uncertainty estimation approaches, Dirichlet Prior Network (DPN) distinctly models different predictive uncertainty types. However, for in-domain examples with high data uncertainties among multiple classes, even a DPN model often produces indistinguishable representations from the out-of-distribution (OOD) examples, compromising their OOD detection performance. We address this shortcoming by proposing a novel loss function for DPN to maximize the \textit{representation gap} between in-domain and OOD examples. Experimental results demonstrate that our proposed approach consistently improves OOD detection performance.
Related Papers
ZClassifier: Temperature Tuning and Manifold Approximation via KL Divergence on Logit Space2025-07-14Safe Domain Randomization via Uncertainty-Aware Out-of-Distribution Detection and Policy Adaptation2025-07-08FA: Forced Prompt Learning of Vision-Language Models for Out-of-Distribution Detection2025-07-06Enclosing Prototypical Variational Autoencoder for Explainable Out-of-Distribution Detection2025-06-17A Variational Information Theoretic Approach to Out-of-Distribution Detection2025-06-17FindMeIfYouCan: Bringing Open Set metrics to $\textit{near} $, $ \textit{far} $ and $\textit{farther}$ Out-of-Distribution Object Detection2025-06-16Optimizing Latent Dimension Allocation in Hierarchical VAEs: Balancing Attenuation and Information Retention for OOD Detection2025-06-11DIsoN: Decentralized Isolation Networks for Out-of-Distribution Detection in Medical Imaging2025-06-10