On Out-of-distribution Detection with Energy-based Models
This work addresses the problem of unreliable OOD detection in machine learning, which is crucial for safety-critical applications, but it is incremental as it builds on prior failures of density estimation methods.
The study investigated whether energy-based models (EBMs) improve out-of-distribution (OOD) detection compared to other density estimation methods, finding that EBMs do not provide consistent advantages and hypothesizing they fail to learn semantic features.
Several density estimation methods have shown to fail to detect out-of-distribution (OOD) samples by assigning higher likelihoods to anomalous data. Energy-based models (EBMs) are flexible, unnormalized density models which seem to be able to improve upon this failure mode. In this work, we provide an extensive study investigating OOD detection with EBMs trained with different approaches on tabular and image data and find that EBMs do not provide consistent advantages. We hypothesize that EBMs do not learn semantic features despite their discriminative structure similar to Normalizing Flows. To verify this hypotheses, we show that supervision and architectural restrictions improve the OOD detection of EBMs independent of the training approach.