Reproduce “Out-of-distribution Detection with Implicit Outlier Transformation” 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, Junjie Ye, Feng Liu, Quanyu Dai, Marcus Kalander, Tongliang Liu, Jianye Hao, Bo Han
DOE uses parameter perturbations to create implicit variations of surrogate outliers, then trains an image classifier to detect out-of-distribution inputs. In a CIFAR-100 comparison using the same MaxLogit score, it lowers average FPR95 from 35.95% to 25.38% and raises AUROC from 92.42% to 93.97% across five OOD test datasets, but does not establish success on arbitrary unseen distributions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.