Research map / Label-efficient and robust learning / Recognition and long-tail
Classification from Positive, Unlabeled and Biased Negative Data
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 |
|---|---|---|---|
| Use PUbN\N instead of nnPU when no biased-negative set is available | PU classification · MNIST, CIFAR-10, and 20 Newsgroups PU tasks | misclassification rate: 5.76 ± 1.04 -> 4.64 ± 0.62; 12.02 ± 0.65 -> 10.70 ± 0.57; 23.78 ± 1.04 -> 21.13 ± 0.90; 14.67 ± 0.87 -> 13.30 ± 0.53 | Helped |
| Use biased-negative data with PUbN rather than PU-only nnPU | PUbN classification · MNIST, CIFAR-10, and 20 Newsgroups tasks with bN | misclassification rate: 5.76 ± 1.04 -> 4.05 ± 0.27; 12.02 ± 0.65 -> 9.71 ± 0.51; 23.78 ± 1.04 -> 18.83 ± 0.71; 14.67 ± 0.87 -> 12.61 ± 0.75 | Helped |
| Increase τ to involve more unlabeled examples as negative evidence (ablation) | PUbN · CIFAR-10, P=vehicles, bN weighted toward horse/deer/frog | misclassification rate: 9.92 -> 11.85 | Hurt |
| Increase τ from 0.7 to 0.9 (ablation) | PUbN · MNIST, P=even digits, bN=1, 3, 5 | misclassification rate: 4.05 -> 3.83 | Helped |
| Misspecify ρ as 1.2ρ instead of ρ (ablation) | PUbN · CIFAR-10, P=vehicles, bN=cat/dog/horse | misclassification rate: 9.71±0.51 -> 10.32±0.57 | Hurt |
Closest papers by meaning
- Balancing Positive and Negative Classification Error Rates in Positive-Unlabeled Learning (NeurIPS 2025 · similarity 0.92)
- Learning from Positive and Unlabeled Data with a Selection Bias (ICLR 2019 · similarity 0.92)
- Accessible, Realistic, and Fair Evaluation of Positive-Unlabeled Learning Algorithms (ICLR 2026 · similarity 0.92)
- Learning from positive and unlabeled examples -Finite size sample bounds (NeurIPS 2025 · similarity 0.89)
- Positive and Unlabeled Learning with Controlled Probability Boundary Fence (ICML 2024 · similarity 0.88)
- Active Learning From Positive and Unlabeled Examples (NeurIPS 2026 · similarity 0.86)
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.