Research map / Bandits and online learning / Online convex optimization and games
Adapting to Online Label Shift with Provable Guarantees
Yong Bai, Yu-Jie Zhang, Peng Zhao, Masashi Sugiyama, Zhi-Hua Zhou
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 OKM hint to ATLAS (ablation) | ATLAS-ADA · Synthetic Lin, N_t=1 | average error (%): 6.28 -> 5.35 | Improved |
| add OKM hint to ATLAS (ablation) | ATLAS-ADA · Synthetic Sin, N_t=1 | average error (%): 6.90 -> 5.25 | Improved |
| add Fwd hint to ATLAS (ablation) | ATLAS-ADA · Synthetic Squ, N_t=100 | average error (%): 4.27 -> 3.33 | Improved |
| add Fwd hint to ATLAS (ablation) | ATLAS-ADA · Synthetic Ber, N_t=100 | average error (%): 4.04 -> 3.19 | Improved |
| add Win hint to ATLAS (ablation) | ATLAS-ADA · Synthetic Squ, N_t=100 | average error (%): 4.27 -> 4.68 | Hurt |
| add Win hint to ATLAS (ablation) | ATLAS-ADA · Synthetic Ber, N_t=100 | average error (%): 4.04 -> 4.29 | Hurt |
| add Peri hint to ATLAS (ablation) | ATLAS-ADA · Synthetic Lin, N_t=100 | average error (%): 5.44 -> 5.38 | No clear effect |
| replace ROGD risk estimator with proposed unbiased estimator (UOGD) | OGD · Real-world ArXiv, Ber, N_t=10 | average error (%): 24.82 -> 21.53 | Improved |
| online updating instead of fixed initial model (UOGD vs FIX) | UOGD · Real-world ArXiv, Ber, N_t=10 | average error (%): 30.63 -> 21.53 | Improved |
| add meta-base ensemble over step sizes (ATLAS vs UOGD) (ablation) | ATLAS · Real-world ArXiv, Ber, N_t=10 | average error (%): 21.53 -> 21.11 | Improved |
| add meta-base ensemble over step sizes (ATLAS vs UOGD) (ablation) | ATLAS · Real-world ArXiv, Lin, N_t=10 | average error (%): 21.04 -> 22.10 | Hurt |
| add OKM hint to ATLAS (ATLAS-ADA vs ATLAS) (ablation) | ATLAS-ADA · Real-world ArXiv, Ber, N_t=10 | average error (%): 21.11 -> 20.58 | Improved |
Closest papers by meaning
- Label Shift Meets Online Learning: Ensuring Consistent Adaptation with Universal Dynamic Regret (CVPR 2025 · similarity 0.94)
- Online Feature Updates Improve Online (Generalized) Label Shift Adaptation (NeurIPS 2024 · similarity 0.88)
- Efficient Non-stationary Online Learning by Wavelets with Applications to Online Distribution Shift Adaptation (ICML 2024 · similarity 0.88)
- Learning from Noisy Labels via Conditional Distributionally Robust Optimization (NeurIPS 2024 · similarity 0.87)
- Improving self-training under distribution shifts via anchored confidence with theoretical guarantees (NeurIPS 2024 · similarity 0.86)
- Protected Test-Time Adaptation via Online Entropy Matching: A Betting Approach (NeurIPS 2024 · 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.
