Research map / Learning theory and algorithms / Optimal transport
Keypoint-Guided Optimal Transport: Models, Algorithms, and Applications
Xiang Gu, Yucheng Yang, Wei Zeng, Jian Sun, Zongben Xu
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 relation preservation instead of distance preservation for keypoint guidance (ablation) | KPG-RL (SH) · Office-31 HDA (1-shot) | accuracy (%): 64.9 -> 79.9 | Improved |
| add mask-based constraint (ablation) | GW · Office-31 HDA (1-shot) | accuracy (%): 35.6 -> 60.8 | Improved |
| add mask-based constraint (ablation) | HOT · Office-31 HDA (1-shot) | accuracy (%): 44.3 -> 55.9 | Improved |
| add mask-based constraint (ablation) | TLB · Office-31 HDA (1-shot) | accuracy (%): 35.7 -> 58.3 | Improved |
| add keypoint-guided relation preservation on top of masked GW (ablation) | GW (w/ mask) · Office-31 HDA (1-shot) | accuracy (%): 60.8 -> 79.6 | Improved |
| solve with Sinkhorn instead of linear programming (ablation) | KPG-RL · Office-31 HDA (1-shot) | accuracy (%): 79.3 -> 79.9 | Improved |
| use KPG-RL (SH) instead of KPG-RL-GW | KPG-RL · Office-31 HDA (1-shot) | accuracy (%): 79.6 -> 79.9 | No clear effect |
| use KPG-RL (SH) instead of SGW | HDA baseline · Office-31 HDA (1-shot) | accuracy (%): 67.3 -> 79.9 | Improved |
| apply GW without keypoint guidance (negative transfer) (ablation) | GW · Office-31 HDA (1-shot) | accuracy (%): 60.6 -> 35.6 | Hurt |
| add KPG-RL (SH) to labeled-target-only SVM, 1-shot (ablation) | kernel SVM · Office-31 HDA (1-shot) | accuracy (%): 60.6 -> 79.9 | Improved |
| add KPG-RL (SH) to labeled-target-only SVM, 3-shot | kernel SVM · Office-31 HDA (3-shot) | accuracy (%): 80.9 -> 87.7 | Improved |
| use source class centers instead of one random point per class as source keypoints (ablation) | KPG-RL (SH) · Office-31 HDA (1-shot) | accuracy (%) | Improved |
Closest papers by meaning
- Keypoint-Guided Optimal Transport with Applications in Heterogeneous Domain Adaptation (NeurIPS 2022 · similarity 0.94)
- Hierarchical Refinement: Optimal Transport to Infinity and Beyond (ICML 2025 · similarity 0.86)
- Improving Neural Optimal Transport via Displacement Interpolation (ICLR 2025 · similarity 0.86)
- Efficient Transferable Optimal Transport via Min-Sliced Transport Plans (NeurIPS 2026 · similarity 0.86)
- Overcoming Spurious Solutions in Semi-Dual Neural Optimal Transport: A Smoothing Approach for Learning the Optimal Transport Plan (ICML 2025 · similarity 0.85)
- Low-Rank Optimal Transport through Factor Relaxation with Latent Coupling (NeurIPS 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.
