Research map / Bandits and online learning / Online convex optimization and games
Confidence Weighted Multitask Learning
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 active adaptive-margin query to CWMT (ACWMT vs CWMT) (ablation) | CWMT · Spam Email | queried labels: 7068 -> 3983.2 | Helped |
| add active adaptive-margin query to CWMT (ablation) | CWMT · MHC-I | queried labels: 18664 -> 11388.5 | Helped |
| add active adaptive-margin query to CWMT (ablation) | CWMT · EachMovie | queried labels: 6000 -> 3171.5 | Helped |
| add active adaptive-margin query to CWMT (ablation) | CWMT · Spam Email, MHC-I, EachMovie | cumulative error rate (%): 4.90 -> 4.89; 36.72 -> 36.73; 17.98 -> 17.97 | No clear effect |
| use second-order confidence-weighted multitask query (ACWMT) instead of first-order ALP query | ALP · Spam Email | cumulative error rate (%): 7.49 -> 4.89 | Helped |
| use ACWMT instead of ALP | ALP · MHC-I | cumulative error rate (%): 42.64 -> 36.73 | Helped |
| use ACWMT instead of ALP | ALP · EachMovie | cumulative error rate (%): 21.64 -> 17.97 | Helped |
| use ACWMT instead of ALP | ALP · Spam Email | queried labels: 5530.8 -> 3983.2 | Helped |
| add second-order confidence weighting (CWMT vs OSMTL-e) | OSMTL-e · Spam Email | cumulative error rate (%): 5.22 -> 4.90 | Helped |
| add second-order confidence weighting (CWMT vs OSMTL-e) | OSMTL-e · EachMovie | cumulative error rate (%): 19.24 -> 17.98 | Helped |
| track covariance of weights (ACWMT vs OSMTL-e) | OSMTL-e · Spam Email | run time (s): 1.36 -> 11.49 | Hurt |
| vary query parameter h across 1e-4 to 1 (ablation) | ACWMT vs ALP · three datasets | cumulative error rate (%) at matched query ratio | Helped |
Closest papers by meaning
- Multitask Learning with No Regret: from Improved Confidence Bounds to Active Learning (NeurIPS 2023 · similarity 0.90)
- Active Learning from Peers (NeurIPS 2017 · similarity 0.88)
- Online Multitask Learning with Long-Term Memory (NeurIPS 2020 · similarity 0.88)
- Adaptive Smoothed Online Multi-Task Learning (NeurIPS 2016 · similarity 0.87)
- When More Data Doesn't Help: Limits of Adaptation in Multitask Learning (ICML 2026 · similarity 0.86)
- Bayesian Uncertainty for Gradient Aggregation in Multi-Task Learning (ICML 2024 · 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.