Research map / Label-efficient and robust learning / Domain adaptation and generalization
DART: Diversify-Aggregate-Repeat Training Improves Generalization of Neural Networks
Samyak Jain, Sravanti Addepalli, Pawan Kumar Sahu, Priyam Dey, R. Venkatesh Babu
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 Mixed Training with DART's diversify-aggregate-repeat training | WideResNet-28-10 · CIFAR-10 | accuracy: 97.69 std 0.19 -> 97.96 std 0.06 | Improved |
| replace Mixed Training with DART's diversify-aggregate-repeat training | WideResNet-28-10 · CIFAR-100 | accuracy: 85.57 std 0.13 -> 86.46 std 0.12 | Improved |
| apply DART to SWAD | ResNet-50 with SWAD · VLCS, PACS, OfficeHome, TerraIncognita, DomainNet | OOD accuracy: 79.1 ± 0.1 -> 80.3 ± 0.2; 88.1 ± 0.1 -> 88.9 ± 0.1; 70.6 ± 0.2 -> 71.9 ± 0.1; 50.0 ± 0.3 -> 51.3 ± 0.2; 46.5 ± 0.1 -> 47 | Improved |
| add DART to MIRO | MIRO with CLIP ViT-B/16 · OfficeHome | OOD accuracy: 83.3 -> 87.37 | Improved |
| use diverse mixed augmentations across branches rather than the same augmentation in every branch (ablation) | DART · CIFAR-100 | accuracy: DART Mixed-Train. 86.13; DART Cutout 84.15 | Improved |
| use DART with the same augmentation in every branch (ablation) | DART · CIFAR-100 | accuracy: ERM + EMA 81.67 -> DART 82.31 | Improved |
| perform interpolation from the beginning rather than after common training (ablation) | DART · CIFAR-100 | accuracy: 0.22% | Hurt |
| increase interpolation interval beyond the useful range (ablation) | DART · CIFAR-100 | accuracy | Hurt |
| increase the number of branches (ablation) | DART · CIFAR-100 | accuracy | Improved |
| delay a single interpolation for progressively longer after common training (ablation) | DART · CIFAR-100 | accuracy: epoch-500 accuracy reaches 0 | Hurt |
| extend training beyond 300 epochs with DART rather than Mixed Training (ablation) | DART vs ERM+EMA Mixed Training · CIFAR-100 | accuracy | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SWAD — SWAD: Domain Generalization by Seeking Flat Minima (NeurIPS 2021) · better on 6 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Seeking Consistent Flat Minima for Better Domain Generalization via Refining Loss Landscapes (CVPR 2025 · similarity 0.86)
- Enhancing Evolving Domain Generalization through Dynamic Latent Representations (AAAI 2024 · similarity 0.86)
- Gradient-Guided Annealing for Domain Generalization (CVPR 2025 · similarity 0.85)
- Improving Generalization via Meta-Learning on Hard Samples (CVPR 2024 · similarity 0.85)
- Sharpness-Aware Model-Agnostic Long-Tailed Domain Generalization (AAAI 2024 · similarity 0.85)
- Simplicity Bias and Optimization Threshold in Two-Layer ReLU Networks (ICML 2025 · similarity 0.85)
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.
