Research map / Label-efficient and robust learning / Semi-supervised learning
Label Propagation with Weak Supervision
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 |
|---|---|---|---|
| Add weak-label aggregation as a prior to LPA | LPA · YouTube, SMS, Basketball, and CDR test sets | accuracy: 82.00 $\pm$ 1.37 -> 88.32 $\pm$ 0.50; 94.32 $\pm$ 0.45 -> 96.80 $\pm$ 0.36; 78.71 $\pm$ 2.41 -> 83.13 $\pm$ 1.43; 67.41 | Helped |
| Apply LPA smoothness to Snorkel weak-label aggregation | Snorkel + L · YouTube test | accuracy: 87.44 $\pm$ 0.47 -> 88.32 $\pm$ 0.50 | Helped |
| Apply LPA smoothness to Snorkel weak-label aggregation | Snorkel + L · SMS test | accuracy: 96.24 $\pm$ 0.32 -> 96.80 $\pm$ 0.36 | Helped |
| Apply LPA smoothness to Snorkel weak-label aggregation | Snorkel + L · Basketball test | accuracy: 82.08 $\pm$ 0.83 -> 83.13 $\pm$ 1.43 | Helped |
| Apply LPA smoothness to Snorkel weak-label aggregation | Snorkel + L · CDR test | accuracy: 68.03 $\pm$ 0.28 -> 67.61 $\pm$ 0.19 | Hurt |
| Use individual weak labelers as accuracy-weighted dongle nodes rather than one aggregated prior | LPA · YouTube test | accuracy: 88.32 $\pm$ 0.50 -> 90.32 $\pm$ 0.43 | Helped |
| Use individual weak labelers as probabilistically weighted dongle nodes rather than one aggregated prior | LPA · CDR test | accuracy: 67.61 $\pm$ 0.19 -> 68.97 $\pm$ 0.51 | Helped |
| Choose accuracy-based versus probabilistic weighting for dongle edges | LPAD · YouTube, SMS, Basketball, and CDR test sets | accuracy: 90.32 $\pm$ 0.43; 96.32 $\pm$ 0.52; 83.06 $\pm$ 0.74; 68.13 $\pm$ 0.74 vs 87.84 $\pm$ 0.53; 96.64 $\pm$ 0.39; 82.01 $\pm | Mixed |
Closest papers by meaning
- Learning from weak labelers as constraints (ICLR 2025 · similarity 0.91)
- Losses over Labels: Weakly Supervised Learning via Direct Loss Construction (AAAI 2023 · similarity 0.91)
- Learning Dependency Structures for Weak Supervision Models (ICML 2019 · similarity 0.90)
- Consistent One-vs-All Losses Robust to Misspecification of the Weak Label Transition Model (NeurIPS 2026 · similarity 0.89)
- Loss factorization, weakly supervised learning and label noise robustness (ICML 2016 · similarity 0.89)
- Learning with Partial-Label and Unlabeled Data: A Uniform Treatment for Supervision Redundancy and Insufficiency (ICML 2024 · similarity 0.89)
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.