Reproduce “Learning to Augment Distributions for Out-of-Distribution Detection” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Qizhou Wang, Zhen Fang, Yonggang Zhang, Feng Liu, Yixuan Li, Bo Han

DAL trains classifiers to detect out-of-distribution (OOD) inputs by augmenting auxiliary OOD examples with feature-space variants from a Wasserstein ball around their distribution, and uses the worst OOD loss within that family. On CIFAR-100, at 50 epochs, DAL lowers FPR95 to 29.68 from OE’s 37.35 across five OOD datasets; this does not establish gains on every dataset or unseen distribution.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.