Unsupervised out-of-distribution (OOD) detection is essential for the reliability ofmachine learning. Inthe literature, existing work has shown that higher-level semantics captured by hierarchical VAEs can be used to detect OOD instances.

Here, by contrast, we propose to explicitly model thehuman-scenecontacts.

Deep neural networks have achieved remarkable advancements in various areas, such as image processing [18, 6] and natural language processing [16, 21].

We now show that w.h.p. over the choice ofJ, the distribution ofh(X) is close to a Gaussian