AI Paper Atlas

Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning

MGDD: A Meta Generator for Fast Dataset Distillation

Songhua Liu, Xinchao Wang

NeurIPS 2023 · Label-efficient and robust learning › Zero-/few-shot and meta-learning · approach: Normalization and unsupervised meta-learning · 0 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 meta-learned generator instead of training the generator from scratch (ablation)MGDD · CUB200 IPC 1, 1000 stepstest accuracy (%): 6.73 -> 7.40Improved
Use meta-learned generator instead of training from scratch (ablation)MGDD · PACS joint IPC 1, 1000 stepstest accuracy (%): 42.73 -> 43.46Improved
Adapt the meta generator with the FRePo objective instead of running FRePo for the same stepsMGDD · CIFAR10 IPC 1, 2000 stepstest accuracy (%): 26.8 -> 42.6Improved
Adapt the meta generator with the DC objective (ablation)MGDD · CIFAR10 IPC 1, 2000 stepstest accuracy (%): 24.7 -> 34.3Improved
Adapt the meta generator with the DM objective (ablation)MGDD · CIFAR10 IPC 1, 2000 stepstest accuracy (%): 25.7 -> 31.4Improved
Evaluate a single-init adapted generator with a different synthetic initialization (ablation)MGDD · CIFAR10 IPC 10test accuracy (%): 60.76 -> 43.27Hurt
Use multi-initialization during adaptation (ablation)MGDD · CIFAR10 IPC 10, different initializationtest accuracy (%): 43.27 -> 60.24Improved
Add 2x down-sampling parameterization to synthetic data (ablation)MGDD · CIFAR10 IPC 1test accuracy (%): 46.26 -> 48.17Improved
Add 2x down-sampling parameterization to synthetic data (ablation)MGDD · CIFAR100 IPC 1test accuracy (%): 24.66 -> 27.76Improved
Train to full convergence instead of 1000 steps, compared with FRePo baselineMGDD · PACS photo domain IPC 10test accuracy (%): 85.40 -> 82.27Hurt
Remove size-embedding channels from generator (ablation)MGDD · CIFAR10 unseen IPCs 20-40test accuracy (%)Hurt
Replace analytical labels with one-hot labels (ablation)MGDD · CIFAR10 IPC 10test accuracy (%)Hurt

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