Research map / Label-efficient and robust learning / Knowledge distillation
Fixed Anchors Are Not Enough: Dynamic Retrieval and Persistent Homology for Dataset Distillation
Muquan Li, Hang Gou, Yingyi Ma, Rongzheng Wang, Ke Qin, Tao He
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 |
|---|---|---|---|
| add Dynamic Retrieval Connection (DRC) alone (ablation) | FADRM · Tiny-ImageNet IPC=10 | top-1 accuracy: 52.8 -> 54.5 | Improved |
| add Persistent Topology Alignment (PTA) alone (ablation) | FADRM · Tiny-ImageNet IPC=10 | top-1 accuracy: 52.8 -> 53.7 | Improved |
| add both DRC and PTA (ablation) | FADRM · Tiny-ImageNet IPC=10 | top-1 accuracy: 52.8 -> 56.2 | Improved |
| add both DRC and PTA (ablation) | FADRM · ImageNet-1K IPC=10 | top-1 accuracy: 50.9 -> 53.2 | Improved |
| add both DRC and PTA (ablation) | FADRM · CIFAR-100 IPC=10 | top-1 accuracy: 67.9 -> 70.3 | Improved |
| add complexity control in DRC retrieval (lambda 0 -> 0.1) (ablation) | DRC · ImageNet-1K IPC=10 | top-1 accuracy: 51.7 -> 52.9 | Improved |
| raise DRC lambda to 1.0 (ablation) | DRC · ImageNet-1K IPC=10 | top-1 accuracy: 52.9 -> 50.8 | Hurt |
| set PTA weight lambda_topo 0.1 -> 0.5 (ablation) | PTA · ImageNet-1K IPC=10 | top-1 accuracy: 51.7 -> 52.5 | Improved |
| raise PTA weight lambda_topo to 3.0 (ablation) | PTA · ImageNet-1K IPC=10 | top-1 accuracy: 52.5 -> 51.1 | Hurt |
| replace FADRM+ with RETA | distilled data · ImageNet-1K IPC=50 ResNet-18 | top-1 accuracy: 61.2 -> 64.3 | Improved |
| replace FADRM+ with RETA | distillation cost · ImageNet-1K | time per image / memory: 1.09s -> 1.31s; 11.0GB -> 12.6GB | Hurt |
Closest papers by meaning
- Homology Consistency Constrained Efficient Tuning for Vision-Language Models (NeurIPS 2024 · similarity 0.86)
- Topology-Aware Representation Alignment for Semi-Supervised Vision-Language Learning (NeurIPS 2026 · similarity 0.86)
- Towards Scalable Topological Regularizers (ICLR 2025 · similarity 0.86)
- Large Scale Dataset Distillation with Domain Shift (ICML 2024 · similarity 0.85)
- Persistence Homology Distillation for Semi-supervised Continual Learning (NeurIPS 2024 · similarity 0.85)
- DELT: A Simple Diversity-driven EarlyLate Training for Dataset Distillation (CVPR 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.
