Research map / Label-efficient and robust learning / Knowledge distillation
Frequency Domain-Based Dataset Distillation
DongHyeok Shin, Seungjae Shin, Il-chul Moon
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 |
|---|---|---|---|
| apply FreD parameterization on top of trajectory matching | TM · CIFAR-10 IPC=2 | test accuracy (%): 50.6 -> 61.4 | Improved |
| apply FreD instead of IDC upsampling | TM · ImageNet-Subset-C IPC=10, ImgNette-C | test accuracy (%): 40.4 -> 55.2 | Improved |
| apply IDC parameterization to TM on corrupted ImageNet-Subset | TM · ImageNet-Subset-C IPC=1, ImgNette-C | test accuracy (%): 38.0 -> 34.5 | Hurt |
| add FreD to HaBa | TM w/ HaBa · CIFAR-10 IPC=2 | test accuracy (%): 61.3 -> 62.3 | Improved |
| use high-pass mask instead of EVR mask (ablation) | FreD · CIFAR-10 IPC=1 | test accuracy (%): 14.32 | Hurt |
| use EVR-based mask instead of low-pass or amplitude-based masks (ablation) | FreD · CIFAR-10 | test accuracy (%) | Improved |
| use DCT instead of DFT or DWT (ablation) | FreD · CIFAR-10 | test accuracy (%) | Improved |
| use more instances with fewer dimensions (small k) at small budget (ablation) | FreD with DM · CIFAR-10 IPC=1 | test accuracy (%) | Improved |
| use more dimensions per instance (large k) at large budget (ablation) | FreD with DM · CIFAR-10 IPC=50 | test accuracy (%) | Improved |
| apply FreD to 3D point clouds with DM | DM · 3D MNIST IPC=1 | test accuracy (%): 42.5 -> 54.9 | Improved |
| apply IDC to DM on 3D MNIST at larger IPC | DM · 3D MNIST IPC=50 | test accuracy (%): 64.7 -> 56.8 | Hurt |
Closest papers by meaning
- Large Scale Dataset Distillation with Domain Shift (ICML 2024 · similarity 0.87)
- Embarrassingly Simple Dataset Distillation (ICLR 2024 · similarity 0.86)
- Towards Lossless Dataset Distillation via Difficulty-Aligned Trajectory Matching (ICLR 2024 · similarity 0.85)
- Boost Self-Supervised Dataset Distillation via Parameterization, Predefined Augmentation, and Approximation (ICLR 2025 · similarity 0.85)
- Group Distributionally Robust Dataset Distillation with Risk Minimization (ICLR 2025 · similarity 0.85)
- Towards Combating Frequency Simplicity-biased Learning for Domain Generalization (NeurIPS 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.
