Research map / Label-efficient and robust learning / Semi-supervised learning
Meta-Learning with Less Forgetting on Large-Scale Non-stationary Task Distributions
Zhenyi Wang, Li Shen, Le Fang, Qiuling Suo, Donglin Zhan, Tiehang Duan, Mingchen Gao
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 MI regularization (ablation) | ORDER · 5-shot, 250 OOD | accuracy: 46.25 -> 50.23 | Improved |
| add OT regularization (labeled features only) (ablation) | ORDER · 5-shot, 250 OOD | accuracy: 46.25 -> 47.69 | Improved |
| add unlabeled ID features to OT loss (ablation) | ORDER (OT) · 5-shot, 250 OOD | accuracy: 47.69 -> 48.57 | Improved |
| add OT on top of MI (ablation) | ORDER · 5-shot, 250 OOD | accuracy: 50.23 -> 51.87 | Improved |
| add OT on top of MI (ablation) | ORDER · 5-shot, 50 OOD | accuracy: 52.63 -> 53.96 | Improved |
| filter OOD data out of unlabeled set (ablation) | ORDER baseline · 5-shot, 250 OOD | accuracy: 46.25 -> 48.36 | Improved |
| add MI on ID data after filtering OOD (ablation) | ORDER · 5-shot, 250 OOD | accuracy: 48.36 -> 49.98 | Improved |
| keep OOD data and add MI with both terms (ablation) | ORDER · 5-shot, 250 OOD | accuracy: 49.98 -> 51.87 | Improved |
| use unlabeled data in meta-learning (MSKM vs ProtoNet) (ablation) | ProtoNet · SETS 5-shot | accuracy: 42.62 -> 45.41 | Improved |
| use unlabeled data (LST vs ANIL) (ablation) | ANIL · SETS 1-shot | accuracy: 30.58 -> 34.81 | Improved |
| add OT regularization with replay vs best CL baseline | Semi-MER · SETS 1-shot | accuracy: 37.90 -> 39.21 | Improved |
| add MI on top of OT (full ORDER vs CL baselines) | Semi-MER · SETS 5-shot | accuracy: 50.46 -> 53.96 | Improved |
Closest papers by meaning
- Meta-Learning without Memorization (ICLR 2020 · similarity 0.88)
- Clustered Task-Aware Meta-Learning by Learning From Learning Paths (TPAMI 2023 · similarity 0.88)
- Secure Out-of-Distribution Task Generalization with Energy-Based Models (NeurIPS 2023 · similarity 0.88)
- Large-Scale Meta-Learning with Continual Trajectory Shifting (ICML 2021 · similarity 0.88)
- Advances and Challenges in Meta-Learning: A Technical Review (TPAMI 2024 · similarity 0.88)
- Task Groupings Regularization: Data-Free Meta-Learning with Heterogeneous Pre-trained Models (ICML 2024 · 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.
