Research map / Label-efficient and robust learning / Recognition and long-tail
RGMIL: Guide Your Multiple-Instance Learning Model with Regressor
Zhaolong Du, Shasha Mao, Yimeng Zhang, Shuiping Gou, Licheng Jiao, Lin Xiong
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 normalization formula to baseline RGP (ablation) | RGP · MMNIST 512/1 | instance-level accuracy (%): 66.54 -> 72.70 | Improved |
| add normalization formula to baseline RGP (ablation) | RGP · MMNIST 64/1 | instance-level accuracy (%): 62.75 -> 90.88 | Improved |
| add difference formula on top of normalization (ablation) | RGP · MMNIST 512/1 | instance-level accuracy (%): 72.70 -> 87.28 | Improved |
| add difference formula on top of normalization (ablation) | RGP · MMNIST 64/1 | instance-level accuracy (%): 90.88 -> 94.50 | Improved |
| add difference formula on top of normalization (ablation) | RGP · MMNIST 10/1 | instance-level accuracy (%): 95.31 -> 98.25 | Improved |
| add normalization formula to baseline RGP (ablation) | RGP · MMNIST 10/1 | instance-level accuracy (%): 94.60 -> 95.31 | No clear effect |
| increase training bag size from 10 to 512 | RGP · MMNIST instance-level | instance-level accuracy (%): 98.25 -> 87.28 | Hurt |
| increase training bag size from 10 to 512 | ABP · MMNIST instance-level | instance-level accuracy (%): 92.18 -> 42.21 | Hurt |
| replace RGP with max pooling aggregator | RGMIL · MMNIST 512/1 | instance-level accuracy (%): 87.28 -> 65.37 | Hurt |
| replace RGP with gated attention pooling (G-ABP) | RGMIL · MMNIST 10/1 | instance-level accuracy (%): 98.25 -> 91.47 | Hurt |
| use RGP instead of attention pooling (ABP) | RGMIL · MMNIST 4-class general | instance-level accuracy (%): 71.76 -> 88.79 | Improved |
| use RGP instead of attention pooling (ABP) | RGMIL · MMNIST 4-class general 512/1 | instance-level accuracy (%): 62.43 -> 81.46 | Improved |
Closest papers by meaning
- Model Agnostic Interpretability for Multiple Instance Learning (ICLR 2022 · similarity 0.85)
- Rethinking Multiple-Instance Learning From Feature Space to Probability Space (ICLR 2025 · similarity 0.85)
- Navigating the MIL Trade-Off: Flexible Pooling for Whole Slide Image Classification (NeurIPS 2025 · similarity 0.84)
- Do Multiple Instance Learning Models Transfer? (ICML 2025 · similarity 0.83)
- Are Multiple Instance Learning Algorithms Learnable for Instances? (NeurIPS 2024 · similarity 0.83)
- How Effective Can Dropout Be in Multiple Instance Learning ? (ICML 2025 · similarity 0.83)
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.
