Research map / Bandits and online learning / Online convex optimization and games
Adaptive Smoothed Online Multi-Task 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 |
|---|---|---|---|
| adaptive loss-based task weights (OSMTL-e) instead of fixed weights (FOML) (ablation) | online multi-task learning · Sentiment Analysis | AUC: 0.8472 -> 0.9465 | Helped |
| adaptive weights (OSMTL-e) instead of fixed weights (FOML) (ablation) | online multi-task learning · Spam Detection | AUC: 0.9347 -> 0.9596 | Helped |
| adaptive weights (OSMTL-t) instead of fixed weights (FOML) (ablation) | online multi-task learning · Landmine Detection | AUC: 0.6322 -> 0.6776 | Helped |
| add loss-based weights to task relationship (OSMTL-e vs OMTRL) | online multi-task learning · Sentiment Analysis | AUC: 0.7831 -> 0.9465 | Helped |
| thresholding updates (OSMTL-t) vs exponential updates (OSMTL-e) (ablation) | OSMTL · Landmine Detection | AUC: 0.6404 -> 0.6776 | Mixed |
| OSMTL-e vs OSGL (no task-relationship learning) (ablation) | online multi-task · Spam Detection | AUC: 0.9551 -> 0.9596 | Helped |
| OSMTL-e vs OSGL (ablation) | online multi-task · Landmine Detection | AUC: 0.6482 -> 0.6404 | Hurt |
| single global model (PA-ONE) vs independent per-task models (PA-ITL) | Passive-Aggressive · Spam Detection | AUC: 0.8350 -> 0.8739 | Helped |
| single global model (PA-ONE) vs independent models (PA-ITL) | Passive-Aggressive · Landmine Detection | AUC: 0.5986 -> 0.5473 | Hurt |
| exponential smoothing updates (OSMTL-e) instead of SVD-based covariance (OMTRL) (ablation) | online multi-task · Spam Detection | CPU time (s): 53.6 -> 1.3 | Helped |
| OSMTL-e vs OMTRL support vectors | online multi-task · Sentiment Analysis | nSV: 1346.2 -> 1322.2 | Helped |
| adaptive weights OSMTL-t vs PA-ITL | online multi-task · Landmine Detection | nSV: 618.1 -> 333.6 | Helped |
Closest papers by meaning
- Multitask Learning with No Regret: from Improved Confidence Bounds to Active Learning (NeurIPS 2023 · similarity 0.87)
- Confidence Weighted Multitask Learning (AAAI 2019 · similarity 0.87)
- Distributed Primal-Dual Optimization for Online Multi-Task Learning (AAAI 2020 · similarity 0.87)
- Parameter-free, Dynamic, and Strongly-Adaptive Online Learning (ICML 2020 · similarity 0.85)
- A Two-Stage Learning-to-Defer Approach for Multi-Task Learning (ICML 2025 · similarity 0.84)
- A Simple and Adaptive Learning Rate for FTRL in Online Learning with Minimax Regret of $\Theta(T^{2/3})$ and its Application to Best-of-Both-Worlds (NeurIPS 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.