AI Paper Atlas

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

Dataset Condensation with Contrastive Signals

Saehyung Lee, Sanghyuk Chun, Sangwon Jung, Sangdoo Yun, Sungroh Yoon

ICML 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Loss design and classifier theory · 9 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
replace class-wise gradient matching with collective, cross-class gradient matchingDataset condensation on fine-grained image classification · Automobile, Terrier, Fish, Lizard, Truck, and Insecttest accuracyImproved
use class-wise gradient matching rather than collective cross-class matchingDataset condensation · Fine-grained Automobile, Terrier, and Fishtest accuracy: 11.0 vs 12.2Hurt
add differentiable Siamese augmentation to DCCDCC · SVHN, CIFAR-10, CIFAR-100 and fine-grained benchmarkstest accuracyMixed
add bi-level warm-up (ablation)DCC · CIFAR-10, 1 image/classtest accuracy: 28.3 -> 32.9Improved
add bi-level warm-up (ablation)DCC · CIFAR-10, 10 images/classtest accuracy: 49.2 -> 49.4Improved
add bi-level warm-up (ablation)DCC · CIFAR-100, 1 image/classtest accuracy: 12.0 -> 13.3Improved
add bi-level warm-up (ablation)DCC · CIFAR-100, 10 images/classtest accuracy: 28.5 -> 30.6Improved
replace no warm-up with simple inner-loop warm-up (ablation)DCC · CIFAR-100, 1 image/classtest accuracy: 11.98 -> 13.01Improved
replace simple warm-up with proposed bi-level warm-up (ablation)DCC · CIFAR-100, 1 image/classtest accuracy: 13.01 -> 13.27Improved
replace no warm-up with simple inner-loop warm-up (ablation)DCC · CIFAR-100, 10 images/classtest accuracy: 28.46 -> 29.74Improved
replace simple warm-up with proposed bi-level warm-up (ablation)DCC · CIFAR-100, 10 images/classtest accuracy: 29.74 -> 30.59Improved

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