AI Paper Atlas

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

Density-driven Regularization for Out-of-distribution Detection

Wenjian Huang, Hao Wang, Jiahao Xia, Chengyan Wang, Jianguo Zhang

NeurIPS 2022 · Label-efficient and robust learning › OOD and anomaly detection · approach: Provable and feature-based OOD detection · 1 citations (OpenAlex)

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
add density consistency regularization (DCR) (ablation)Ours · CIFAR-100 (WideResNet)mean FPR95%: 77.04 -> 41.59Improved
add contrastive distribution regularization (CDR) on top of DCR (ablation)Ours · CIFAR-100 (WideResNet)mean FPR95%: 41.59 -> 13.35Improved
add DCR to baseline (ablation)Ours · CIFAR-100 (WideResNet)mean AUROC: 80.05 -> 90.84Improved
add CDR on top of DCR (ablation)Ours · CIFAR-100 (WideResNet)mean AUROC: 90.84 -> 96.09Improved
batch size 256 -> 128 (ablation)Ours · CIFAR-100 (WideResNet)mean FPR95%: 13.35 -> 13.43No clear effect
batch size 256 -> 512 (ablation)Ours · CIFAR-100 (WideResNet)mean FPR95%: 13.35 -> 13.63No clear effect
regularization weight gamma 0.01 -> 0.005 (ablation)Ours · CIFAR-100 (WideResNet)mean FPR95%: 13.35 -> 14.08Mixed
regularization weight gamma 0.01 -> 0.02 (ablation)Ours · CIFAR-100 (WideResNet)mean FPR95%: 13.35 -> 12.92No clear effect
constant r 10 -> 5 (ablation)Ours · CIFAR-100 (WideResNet)mean FPR95%: 13.29 -> 14.03No clear effect
constant r 10 -> 20 (ablation)Ours · CIFAR-100 (WideResNet)mean FPR95%: 13.29 -> 12.87No clear effect
use proposed method instead of Gram on average FPR95%Ours · CIFAR-100 (ResNet34)mean FPR95%: 26.51 -> 16.60Improved
use proposed method instead of Gram on average FPR95%Ours · CIFAR-100 (WideResNet)mean FPR95%: 27.66 -> 13.35Improved

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