Research map / Label-efficient and robust learning / Recognition and long-tail
Loss-Based Attention for Deep Multiple Instance Learning
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 |
|---|---|---|---|
| loss-based attention instead of Attention MIL pooling | MI-Net architecture · FOX | accuracy: 0.615±0.096 -> 0.712±0.074 | Helped |
| loss-based attention instead of Attention+softmax | MIL network · MNIST-bags binary 50 bags | AUC: 0.914 -> 0.931 | Helped |
| loss-based attention instead of Attention+softmax | MIL network · MNIST-bags multi-class 150 bags | accuracy: 0.923 -> 0.917 | Hurt |
| loss-based attention instead of Attention+softmax | MIL network · CIFAR10-bags | accuracy: 0.823 -> 0.839 | Helped |
| replace global mean-pooling with loss-based attention | ResNet18 · CIFAR-10 1000 images | accuracy: 0.584 -> 0.605 | Helped |
| replace global mean-pooling with loss-based attention | ResNet18 · Tiny ImageNet | top-5 accuracy: 0.791 -> 0.785 | Hurt |
| replace global mean-pooling with loss-based attention | ResNet18 · Tiny ImageNet | localization AP: 0.704 -> 0.756 | Helped |
| add recall regularization term L2 (lambda > 0) (ablation) | Loss-Attention · MNIST-bags 50 bags | instance F-score | Helped |
| increase lambda (ablation) | Loss-Attention · MNIST-bags 50 bags | instance precision vs recall | Mixed |
Closest papers by meaning
- Attention-based Deep Multiple Instance Learning (ICML 2018 · similarity 0.90)
- Are Multiple Instance Learning Algorithms Learnable for Instances? (NeurIPS 2024 · similarity 0.88)
- Multi-Instance Causal Representation Learning for Instance Label Prediction and Out-of-Distribution Generalization (NeurIPS 2022 · similarity 0.85)
- Multiple Instance Verification (NeurIPS 2026 · similarity 0.84)
- Boosting Multiple Instance Learning Models for Whole Slide Image Classification: A Model-Agnostic Framework Based on Counterfactual Inference (AAAI 2024 · similarity 0.84)
- Forming Auxiliary High-confident Instance-level Loss to Promote Learning from Label Proportions (CVPR 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.