AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

On Predicting Generalization using GANs

Yi Zhang, Arushi Gupta, Nikunj Saunshi, Sanjeev Arora

ICLR 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Data augmentation and generalization bias · 2 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
Use GAN-generated labeled samples to predict classifier generalizationDeep classifiers on PGDL tasks · PGDL Task 1: VGG on CIFAR-10Conditional Mutual Information: 52.59 -> 62.62Improved
Use GAN-generated labeled samples to predict classifier generalizationDeep classifiers on PGDL tasks · PGDL Task 2: NIN on SVHNConditional Mutual Information: 48.34 -> 34.72Hurt
Use GAN-generated labeled samples to predict classifier generalizationDeep classifiers on PGDL tasks · PGDL Task 4: AllConv on CINIC-10Conditional Mutual Information: 57.81 -> 52.80Hurt
Use GAN-generated labeled samples to predict classifier generalizationDeep classifiers on PGDL tasks · PGDL Task 5: AllConv on CINIC-10Conditional Mutual Information: 34.57 -> 53.56Improved
Use GAN-generated labeled samples to predict classifier generalizationDeep classifiers on PGDL tasks · PGDL Task 8: VGG on F-MNISTConditional Mutual Information: 16.23 -> 30.25Improved
Use GAN-generated labeled samples to predict classifier generalizationDeep classifiers on PGDL tasks · PGDL Task 9: NIN on CIFAR-10Conditional Mutual Information: 25.86 -> 33.51Improved
Use BigGAN+DiffAug synthetic examples rather than the qrt+log predictorDEMOGEN generalization prediction · NiN+CIFAR-10adjusted R²: 0.94 -> 0.995Improved
Use BigGAN+DiffAug synthetic examples rather than the qrt+log predictorDEMOGEN generalization prediction · ResNet+CIFAR-10adjusted R²: 0.87 -> 0.981Improved
Use BigGAN+DiffAug synthetic examples rather than the qrt+log predictorDEMOGEN generalization prediction · ResNet+CIFAR-100adjusted R²: 0.97 -> 0.993Improved
Apply ordinary data augmentation to real GAN-training examples only (ablation)BigGAN generalization predictor · DEMOGEN CIFAR-100test accuracy versus synthetic accuracy agreementHurt
Remove data augmentation from GAN training (ablation)BigGAN generalization predictor · DEMOGEN CIFAR-100test accuracy versus synthetic accuracy agreementImproved
Apply differentiable augmentation to both real and fake GAN images (ablation)BigGAN generalization predictor · DEMOGEN CIFAR-100test accuracy versus synthetic accuracy agreementImproved

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