Research map / Graph learning / Graph algorithms and clustering
Selecting Samples on Graphs: A Unified Dataset Pruning Framework for Lossless Training Acceleration
Dongyue Wu, Zilin Guo, Xiaoyu Li, Jiajia Liu, Jingdong Chen, Nong Sang, Changxin Gao
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 the extrinsic (diversity) term to the intrinsic term (ablation) | UGIES · ImageNet-1k ResNet-50 | accuracy | Improved |
| use greedy solver instead of top-k selection (ablation) | UGIES · ImageNet-1k ResNet-50 | accuracy | Improved |
| promote diversity (monotone-increasing extrinsic term) instead of similarity (ablation) | UGIES · ImageNet-1k ResNet-50 | accuracy | Improved |
| use cosine distance instead of Chebyshev distance (ablation) | UGIES · ImageNet-1k ResNet-50 (30% pruned) | accuracy: 76.6 -> 77.0 | Improved |
| use cosine distance instead of L2 distance (ablation) | UGIES · ImageNet-1k ResNet-50 (30% pruned) | accuracy: 76.8 -> 77.0 | Improved |
| change the mapping function g among g1 to g4 (ablation) | UGIES · ImageNet-1k ResNet-50 (30% pruned) | accuracy: 76.9 -> 77.0 | No clear effect |
| extract features with current training model instead of pretrained ResNet-34 (ablation) | UGIES · ImageNet-1k ResNet-50 (30% pruned) | top-1 accuracy: 76.82 -> 77.02 | Improved |
| use pretrained ResNet-50 features instead of ResNet-34 (ablation) | UGIES · ImageNet-1k ResNet-50 (30% pruned) | top-1 accuracy: 76.82 -> 76.96 | No clear effect |
| swap in DivBS gradient similarity as extrinsic term (ablation) | UGIES · ImageNet-1k ResNet-50 (70% pruned) | accuracy: 75.3 -> 75.5 | Improved |
| use UGIES instead of InfoBatch | UGIES · ImageNet-1k ResNet-50 (70% pruned) | accuracy: 74.9 -> 75.3 | Improved |
| use UGIES instead of PFB for training acceleration | UGIES · ImageNet-1k ResNet-50 (50% pruned) | total time (h): 8.7 -> 7.9 | Improved |
| use UGIES instead of InfoBatch | UGIES · COCO2017 SSD ResNet-50 | AP: 25.0 -> 25.5 | Improved |
Closest papers by meaning
- GDeR: Safeguarding Efficiency, Balancing, and Robustness via Prototypical Graph Pruning (NeurIPS 2024 · similarity 0.90)
- Exploring Learning Complexity for Efficient Downstream Dataset Pruning (ICLR 2025 · similarity 0.90)
- $\mathbb{D}^2$ Pruning: Message Passing for Balancing Diversity & Difficulty in Data Pruning (ICLR 2024 · similarity 0.89)
- Lightweight Dataset Pruning without Full Training via Example Difficulty and Prediction Uncertainty (ICML 2025 · similarity 0.88)
- Distilling the Knowledge in Data Pruning (ICML 2025 · similarity 0.88)
- Data Pruning by Information Maximization (ICLR 2025 · similarity 0.87)
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.
