Research map / Graph learning / Graph algorithms and clustering
Revocable Deep Reinforcement Learning with Affinity Regularization for Outlier-Robust Graph Matching
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 affinity regularization to RGM | RGM · Willow Object, 3 outliers | F1: 73.41% -> 75.65% | Helped |
| Add inlier count information to RGM | RGM · Willow Object, 3 outliers | F1: 73.41% -> 76.72% | Helped |
| Combine affinity regularization and inlier count information | RGM · Willow Object, 3 outliers | F1: 75.65% -> 77.57% | Helped |
| Add the revocable action mechanism (ablation) | RGM + AR · Willow Object, 3 outliers | F1: 72.19% -> 75.65% | Helped |
| Add the revocable action mechanism (ablation) | RGM + IC · Willow Object, 3 outliers | F1: 72.42% -> 76.72% | Helped |
| Add the revocable action mechanism (ablation) | RGM + AR + IC · Willow Object, 3 outliers | F1: 73.92% -> 77.57% | Helped |
| Use an inaccurate inlier-count input of 8 instead of the ground-truth count (ablation) | RGM + IC · Willow Object, 3 outliers | F1: 76.72% -> 70.78% | Hurt |
| Use NGM-v2 as guidance for RGM imitation learning | NGM-v2 · Pascal VOC | F1: 80.10% -> 81.88% | Helped |
| Use RGM as the back-end solver | Graph matching on Pascal VOC · Pascal VOC | objective score: 1.040 | Mixed |
| Use RGM instead of NGM | QAP solver · QAPLIB, average | optimal gap: 62.4 -> 35.8 | Helped |
| Use RGM instead of NGM | QAP solver · QAPLIB, average | inference time per instance: 15.72 -> 75.53 | Hurt |
Closest papers by meaning
- Deep Learning of Partial Graph Matching via Differentiable Top-K (CVPR 2023 · similarity 0.89)
- IA-GM: A Deep Bidirectional Learning Method for Graph Matching (AAAI 2021 · similarity 0.88)
- Online Matching with Stochastic Rewards: Provable Better Bound via Adversarial Reinforcement Learning (ICML 2024 · similarity 0.85)
- SiT: Symmetry-invariant Transformers for Generalisation in Reinforcement Learning (ICML 2024 · similarity 0.85)
- Learning Partial Graph Matching via Optimal Partial Transport (ICLR 2025 · similarity 0.85)
- DiMa: Understanding the Hardness of Online Matching Problems via Diffusion Models (ICML 2025 · 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.