Energy Correction Model in the Feature Space for Out-of-Distribution Detection
This addresses the problem of detecting out-of-distribution samples for deep learning systems, but it is incremental as it builds on existing energy-based models and benchmarks.
The paper tackled out-of-distribution detection by using an energy-based model to learn in-distribution features, but found that MCMC sampling issues hurt performance, so they proposed an energy correction method that achieved favorable results compared to strong baselines on CIFAR benchmarks.
In this work, we study the out-of-distribution (OOD) detection problem through the use of the feature space of a pre-trained deep classifier. We show that learning the density of in-distribution (ID) features with an energy-based models (EBM) leads to competitive detection results. However, we found that the non-mixing of MCMC sampling during the EBM's training undermines its detection performance. To overcome this an energy-based correction of a mixture of class-conditional Gaussian distributions. We obtains favorable results when compared to a strong baseline like the KNN detector on the CIFAR-10/CIFAR-100 OOD detection benchmarks.