Research map / Label-efficient and robust learning / Recognition and long-tail
FAST: Topology-Aware Frequency-Domain Distribution Matching for Coreset Selection
Jin Cui, Boran Zhao, Jiajun Xu, Jiaqi Guo, Shuo Guan, Pengju Ren
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 |
|---|---|---|---|
| use phase-decoupled CFD instead of standard CFD (ablation) | FAST · DTD | accuracy (%): 41.15 -> 45.77 | Improved |
| use phase-decoupled CFD instead of standard CFD (ablation) | FAST · RESISC-45 | accuracy (%): 82.01 -> 85.00 | Improved |
| use phase-decoupled CFD instead of standard CFD (ablation) | FAST · CIFAR-10 | accuracy (%): 88.17 -> 90.32 | Improved |
| use PDAS instead of collinear frequency set (ablation) | FAST · CIFAR-10 (10% keep) | accuracy (%): 74.16 -> 90.32 | Improved |
| use PDAS instead of uniform sampling (ablation) | FAST · CIFAR-10 (10% keep) | accuracy (%): 86.60 -> 90.32 | Improved |
| use graph spectral features instead of pixel-level optimization (ablation) | FAST · CIFAR-10 (10% keep) | accuracy (%): 61.52 -> 90.32 | Improved |
| use graph spectral features instead of ResNet-50 features (ablation) | FAST · CIFAR-10 (10% keep) | accuracy (%): 81.51 -> 90.32 | Improved |
| use neighbor scale k=15 instead of k=5 (ablation) | FAST · CIFAR-10 (10% keep) | accuracy (%): 72.12 -> 90.32 | Improved |
| use reduced dimension d=32 instead of d=8 (ablation) | FAST · CIFAR-10 (10% keep) | accuracy (%): 78.86 -> 90.32 | Improved |
| use reduced dimension d=32 instead of d=128 (ablation) | FAST · CIFAR-10 (10% keep) | accuracy (%): 86.73 -> 90.32 | Improved |
| use PD-CFD instead of MSE as the distribution metric (ablation) | FAST · CIFAR-10 (10% keep, ResNet18) | accuracy (%): 9.16% accuracy drop | Improved |
| use PD-CFD instead of KL as the distribution metric (ablation) | FAST · CIFAR-10 (10% keep, ResNet18) | accuracy (%): 7.15% accuracy drop | Improved |
Closest papers by meaning
- Feature Distribution Matching by Optimal Transport for Effective and Robust Coreset Selection (AAAI 2024 · similarity 0.89)
- Refined Coreset Selection: Towards Minimal Coreset Size under Model Performance Constraints (ICML 2024 · similarity 0.89)
- Stable Coresets via Posterior Sampling: Aligning Induced and Full Loss Landscapes (NeurIPS 2025 · similarity 0.88)
- $\mathbb{D}^2$ Pruning: Message Passing for Balancing Diversity & Difficulty in Data Pruning (ICLR 2024 · similarity 0.86)
- Dataset Distillation with Neural Characteristic Function: A Minmax Perspective (CVPR 2025 · similarity 0.86)
- Contributing Dimension Structure of Deep Feature for Coreset Selection (AAAI 2024 · 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.
