Research map / Label-efficient and robust learning / Semi-supervised learning
DivAug: Plug-In Automated Data Augmentation With Explicit Diversity Maximization
Zirui Liu, Haifeng Jin, Ting-Hsiang Wang, Kaixiong Zhou, Xia Hu
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 changed | Where | Measured | Effect |
|---|---|---|---|
| replace k-means++ selection with random selection of augmented images (i.e. add k-means++ selection) (ablation) | DivAug (WRN-28-10) · CIFAR-10 | accuracy: 97.7 -> 98.1 | Improved |
| replace k-means++ selection with random selection (i.e. add k-means++ selection) (ablation) | DivAug (WRN-28-10) · CIFAR-100 | accuracy: 83.3 -> 84.2 | Improved |
| use DivAug instead of RandAugment as perturbation in UDA | UDA · CIFAR-10 SSL (1000 labels) | error rate: 7.37 -> 6.94 | Improved |
| use DivAug instead of RandAugment as perturbation in UDA | UDA · CIFAR-10 SSL (2000 labels) | error rate: 6.50 -> 6.26 | Improved |
| use DivAug instead of RandAugment as perturbation in UDA | UDA · CIFAR-10 SSL (4000 labels) | error rate: 5.44 -> 5.40 | No clear effect |
| use DivAug instead of RandAugment | Wide-ResNet-28-10 · CIFAR-10 | accuracy: 97.3 -> 98.1 | Improved |
| use DivAug instead of Fast AutoAugment | Wide-ResNet-28-10 · CIFAR-10 | accuracy: 97.3 -> 98.1 | Improved |
| use DivAug instead of AutoAugment | Wide-ResNet-40-2 · CIFAR-100 | accuracy: 79.3 -> 81.3 | Improved |
| use DivAug instead of Adversarial AutoAugment | ResNet-50 · ImageNet val | accuracy: 79.4 -> 78.0 | Hurt |
| use DivAug instead of Adversarial AutoAugment | Wide-ResNet-28-10 · CIFAR-100 | accuracy: 84.5 -> 84.2 | Hurt |
| use DivAug instead of Adversarial AutoAugment | Wide-ResNet-28-10 · CIFAR-10 | accuracy: 98.1 -> 98.1 | No clear effect |
| train DivAug with E=8, S=4 instead of RA sampling: extra training cost | DivAug · CIFAR-10 | training cost relative to RA: 1.0 -> 4.5 | Mixed |
Closest papers by meaning
- Tradeoffs in Data Augmentation: An Empirical Study (ICLR 2021 · similarity 0.87)
- The good, the bad and the ugly sides of data augmentation: An implicit spectral regularization perspective (JMLR 2024 · similarity 0.87)
- Characterizing the Generalization Error of Random Feature Regression with Arbitrary Data-Augmentation (NeurIPS 2026 · similarity 0.86)
- When Dynamic Data Selection Meets Data Augmentation: Achieving Enhanced Training Acceleration (ICML 2025 · similarity 0.85)
- Data Augmentation with Diffusion for Open-Set Semi-Supervised Learning (NeurIPS 2024 · similarity 0.85)
- DiverGen: Improving Instance Segmentation by Learning Wider Data Distribution with More Diverse Generative Data (CVPR 2024 · similarity 0.84)
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.
