Research map / Label-efficient and robust learning / Recognition and long-tail
Binary Classification from Multiple Unlabeled Datasets via Surrogate Set Classification
Nan Lu, Shida Lei, Gang Niu, Issei Sato, Masashi Sugiyama
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 surrogate set classification with the transition transformation instead of empirical proportion risk | U^m classification · MNIST, Fashion-MNIST, Kuzushiji-MNIST, CIFAR-10 | test classification error | Improved |
| Use surrogate set classification rather than paired risk-consistent estimators | U^m classification · MNIST, Fashion-MNIST, Kuzushiji-MNIST, CIFAR-10 | test classification error | Improved |
| Use unbiased risk estimators instead of balanced risk estimators | MMC baseline · MNIST, Fashion-MNIST, Kuzushiji-MNIST, CIFAR-10 | test classification error | Improved |
| Apply non-negative risk correction | MMC-U^2 · U^m classification | overfitting | Improved |
| Reduce set size using tau shifts (ablation) | U^m-SSC · MNIST, Fashion-MNIST, Kuzushiji-MNIST, CIFAR-10 | test classification error | Mixed |
| Randomly vary U-set sizes (ablation) | U^m-SSC · MNIST, Fashion-MNIST, Kuzushiji-MNIST, CIFAR-10 | test classification error | Improved |
| Increase the number of U sets from 2 to 100, 500, or 1000 (ablation) | U^m-SSC · MNIST, Fashion-MNIST, Kuzushiji-MNIST, CIFAR-10 | test classification error | Mixed |
| Use noisy class priors instead of true class priors (ablation) | U^m-SSC · 50 U sets | test classification error | Hurt |
Closest papers by meaning
- Making Binary Classification From Multiple Unlabeled Datasets Almost Free of Supervision (TPAMI 2026 · similarity 0.88)
- On the Minimal Supervision for Training Any Binary Classifier from Only Unlabeled Data (ICLR 2019 · similarity 0.87)
- Consistent Multi-Class Classification from Multiple Unlabeled Datasets (ICLR 2024 · similarity 0.85)
- AUC Optimization from Multiple Unlabeled Datasets (AAAI 2024 · similarity 0.84)
- Cost-Sensitive Learning with Noisy Labels (JMLR 2018 · similarity 0.83)
- Accessible, Realistic, and Fair Evaluation of Positive-Unlabeled Learning Algorithms (ICLR 2026 · 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.
