AI Paper Atlas

Research map / Label-efficient and robust learning / OOD and anomaly detection

Out-of-distribution Detection Learning with Unreliable Out-of-distribution Sources

Haotian Zheng, Qizhou Wang, Zhen Fang, Xiaobo Xia, Feng Liu, Tongliang Liu, Bo Han

NeurIPS 2023 · Label-efficient and robust learning › OOD and anomaly detection · approach: Provable and feature-based OOD detection

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use ATOL instead of ManifoldGANdata generation-based OOD detection · CIFAR-10FPR95: 26.68 -> 14.66Improved
use ATOL instead of ManifoldGANdata generation-based OOD detection · ImageNetFPR95: 72.50 -> 51.80Improved
add threshold tau to separate auxiliary ID and OOD latent distributions (ablation)ATOL · CIFAR-100FPR95: 92.90 -> 55.06Improved
add distance-preserving regularization l_reg on the generator (ablation)ATOL · CIFAR-10FPR95: 41.13 -> 22.29Improved
add ID distribution alignment loss l_align (ablation)ATOL · CIFAR-100FPR95: 74.20 -> 55.06Improved
replace DCGAN with BigGAN generator (ablation)ATOL · CIFAR-100FPR95: 55.52 -> 36.72Improved
replace trained DCGAN with random-parameterized DCGAN (ablation)ATOL · CIFAR-100FPR95: 55.52 -> 65.69Hurt
use class-conditional instead of class-agnostic auxiliary ID data (ablation)ATOL · CIFAR-100FPR95: 61.54 -> 55.52Improved
use Euclidean instead of cosine-based centralized distance in l_reg (ablation)ATOL · CIFAR-10FPR95: 19.73 -> 14.62Improved
increase alignment weight alpha from 0.01 to 1 (ablation)ATOL · CIFAR-10FPR95: 22.99 -> 14.11Improved
increase alpha from 1 to 10 (ablation)ATOL · CIFAR-100FPR95: 55.38 -> 61.82Mixed
increase latent space size u from 0.5 to 8 (ablation)ATOL · CIFAR-10FPR95: 18.15 -> 13.90Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy