Research map / Label-efficient and robust learning / Noisy labels
Learning from Label Proportions by Learning with Label Noise
Jianxin Zhang, Yutong Wang, Clayton Scott
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 |
|---|---|---|---|
| Replace KL loss with LLPFC-uniform | Wide ResNet-16-4 · CIFAR10, bag size 256 | test accuracy: .3749 $\pm$ .14 -> .7381 $\pm$ .01 | Improved |
| Replace KL loss with LLPFC-approx | Wide ResNet-16-4 · EMNIST, bag size 512 | test accuracy: .0845 $\pm$ .01 -> .9048 $\pm$ .00 | Improved |
| Replace KL loss with LLPFC-approx | ResNet18 · CIFAR10, bag size 32 | test accuracy: .7837 $\pm$ .01-> .6567 $\pm$ .01 | Hurt |
| Replace KL loss with LLPFC-approx | ResNet18 · CIFAR10, bag size 1024 | test accuracy: .5197 $\pm$ .04 -> .6169 $\pm$ .01 | Improved |
| Replace KL loss with LLPFC-approx | ResNet18 · EMNIST, bag size 32 | test accuracy: .9319 $\pm$ .00 -> .9146 $\pm$ .00 | Hurt |
| Replace KL loss with LLPFC-approx | ResNet18 · SVHN, bag size 32 | test accuracy: .1279 $\pm$ .06 -> .8824 $\pm$ .01 | Improved |
| Replace KL loss with LLPFC-uniform | VGG16 · CIFAR10, bag size 32 | test accuracy: .2513 $\pm$ .11 -> .7602 $\pm$ .00 | Improved |
| Replace KL loss with LLPFC-uniform | VGG16 · SVHN, bag size 256 | test accuracy: .1024 $\pm$ .05 -> .1734 $\pm$ .11 | No clear effect |
| Replace LLPGAN with LLPFC-approx | LLPGAN discriminator · CIFAR10, bag size 32 | test accuracy: .3630 $\pm$ .01 -> .6169 $\pm$ .01 | Improved |
| Replace LLPGAN with LLPFC-uniform | LLPGAN discriminator · SVHN, bag size 32 | test accuracy: .2378 $\pm$ .24 -> .8800 $\pm$ .00 | Improved |
| Use LLPFC-approx instead of LLPFC-uniform | Wide ResNet-16-4 · EMNIST, bag size 2048 | test accuracy: .8817 $\pm$ .01 -> .9007 $\pm$ .01 | Improved |
| Use LLPFC-approx instead of LLPFC-uniform | VGG16 · SVHN, bag size 128 | test accuracy: .5402 $\pm$ .21 -> .1091 $\pm$ .04 | Hurt |
Closest papers by meaning
- Forming Auxiliary High-confident Instance-level Loss to Promote Learning from Label Proportions (CVPR 2025 · similarity 0.93)
- Easy Learning from Label Proportions (NeurIPS 2023 · similarity 0.92)
- Rethinking Learning from Label Proportions via Moment Matching (NeurIPS 2026 · similarity 0.91)
- Learning from Label Proportions: Bootstrapping Supervised Learners via Belief Propagation (ICLR 2024 · similarity 0.90)
- Nearly Optimal Sample Complexity for Learning with Label Proportions (ICML 2025 · similarity 0.89)
- Learning from Noisy Labels via Conditional Distributionally Robust Optimization (NeurIPS 2024 · 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.
