Research map / Learning theory and algorithms / Prediction and decision losses
Mixture Proportion Estimation and PU Learning:A Modern Approach
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 (TED)^n (iterated estimate and discard) instead of BBE for mixture proportion estimation | MPE · Binarized CIFAR (ResNet) | absolute MPE error: 0.091 -> 0.026 | Helped |
| Use (TED)^n instead of BBE for MPE | MPE · CIFAR Dog vs Cat (ResNet) | absolute MPE error: 0.176 -> 0.078 | Helped |
| Use (TED)^n instead of BBE for MPE | MPE · MNIST17 (MLP) | absolute MPE error: 0.023 -> 0.003 | Helped |
| Use (TED)^n instead of BBE for MPE | MPE · Binarized CIFAR (MLP) | absolute MPE error: 0.177 -> 0.225 | Hurt |
| Use (TED)^n instead of BBE for MPE | MPE · Binarized CIFAR (All Conv) | absolute MPE error: 0.037 -> 0.042 | Hurt |
| Use CVIR objective instead of PvU training for PU classification | PU classifier · Binarized CIFAR (ResNet) | accuracy: 76.9 -> 82.3 | Helped |
| Use CVIR instead of PvU training | PU classifier · CIFAR Dog vs Cat (ResNet) | accuracy: 67.3 -> 73.3 | Helped |
| Use CVIR instead of nnPU | PU classifier · Binarized CIFAR (All Conv) | accuracy: 73.4 -> 78.1 | Helped |
| Use CVIR instead of nnPU | PU classifier · Binarized MNIST (MLP) | accuracy: 96.1 -> 96.3 | Helped |
| Use (TED)^n with unknown alpha instead of CVIR with known alpha | PU classifier · Binarized CIFAR (ResNet) | accuracy: 82.3 -> 82.7 | Helped |
| Use (TED)^n with unknown alpha instead of CVIR with known alpha | PU classifier · Binarized CIFAR (MLP) | accuracy: 66.9 -> 64.2 | Hurt |
| Use (TED)^n instead of DEDPUL | PU classifier · IMDb (BERT) | accuracy: 87.3 -> 87.6 | Helped |
Closest papers by meaning
- Rethinking Class-Prior Estimation for Positive-Unlabeled Learning (ICLR 2022 · similarity 0.88)
- Learning from Positive and Unlabeled Data without Explicit Estimation of Class Prior (AAAI 2020 · similarity 0.88)
- PU-BENCH: A UNIFIED BENCHMARK FOR RIGOROUS AND REPRODUCIBLE PU LEARNING (ICLR 2026 · similarity 0.88)
- Mixture Proportion Estimation via Kernel Embeddings of Distributions (ICML 2016 · similarity 0.88)
- Learning from Positive and Unlabeled Data with Arbitrary Positive Shift (NeurIPS 2020 · similarity 0.88)
- Learning from positive and unlabeled examples -Finite size sample bounds (NeurIPS 2025 · similarity 0.88)
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.