AI Paper Atlas

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

Revealing the Distributional Vulnerability of Discriminators by Implicit Generators

Zhilin Zhao, Longbing Cao, Kun-Yu Lin

TPAMI 2022 · Label-efficient and robust learning › OOD and anomaly detection · approach: Provable and feature-based OOD detection · 12 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
fine-tune discriminator on its own implicit-generator OOD samples (FIG)ResNet18 baseline detector · SVHN, CIFAR10, CIFAR100AUROC (%): 1.22% to 23.49% improvementImproved
replace Gaussian-noise OOD samples with implicit-generator samplesGS fine-tuning · SVHN, CIFAR10, CIFAR100AUROC (%): 5.69% improvementImproved
fine-tune with FIG on higher-resolution dataResNet18 · MiniImageNetAUROC (%): 7.87% average AUROC improvementImproved
replace DCC* with FIGResNet18 · MiniImageNetAUROC (%): 81.4 -> 84.4Improved
replace DCC* with FIGResNet18 · CIFAR100AUROC (%): 88.7 -> 89.3Improved
add FIG fine-tuning to ODIN detectorResNet18 · CIFAR100AUROC (%): 84.6 -> 90.2Improved
add FIG fine-tuning to DRF detectorResNet18 · CIFAR100AUROC (%): 71.2 -> 74.4Improved
use FIG instead of JCL / DCC* for balance of ID accuracy and OOD detectionResNet18 · CIFAR10harmonic mean: 46.8 -> 47.5Improved
use adversarial samples (AD) instead of standard trainingResNet18 · CIFAR100harmonic mean: 39.9 -> 35.9Hurt
raise low-entropy percentage K above 0.1 (ablation)FIG · SVHN, CIFAR10, CIFAR100AUROC (%)Hurt
raise low-entropy percentage K from 0 (no generated samples) up to 0.1 (ablation)FIG · SVHN, CIFAR10, CIFAR100AUROC (%)Improved

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