English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Handling out-of-distribution (OOD) samples has become a major stake in thereal-world deployment of machine learning systems. This work explores the useof self-supervised contrastive learning to the simultaneous detection of twotypes of OOD samples: unseen classes and adversarial perturbations. First, wepair self-supervised contrastive learning with the maximum mean discrepancy(MMD) two-sample test. This approach enables us to robustly test whether twoindependent sets of samples originate from the same distribution, and wedemonstrate its effectiveness by discriminating between CIFAR-10 and CIFAR-10.1with higher confidence than previous work. Motivated by this success, weintroduce CADet (Contrastive Anomaly Detection), a novel method for OODdetection of single samples. CADet draws inspiration from MMD, but leveragesthe similarity between contrastive transformations of a same sample. CADetoutperforms existing adversarial detection methods in identifying adversariallyperturbed samples on ImageNet and achieves comparable performance to unseenlabel detection methods on two challenging benchmarks: ImageNet-O andiNaturalist. Significantly, CADet is fully self-supervised and requires neitherlabels for in-distribution samples nor access to OOD examples.

CADet: Fully Self-Supervised Out-Of-Distribution Detection With Contrastive Learning