Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Learning to Forget for Meta-Learning
Check 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 task- and layer-dependent attenuation to MAML | MAML · miniImageNet 5-way 5-shot | accuracy: 63.11 \pm 0.91% -> 69.38 \pm 0.46% | Helped |
| add task- and layer-dependent attenuation to MAML | MAML · tieredImageNet 5-way 5-shot | accuracy: 67.48 \pm 0.47% -> 73.34 \pm 0.44% | Helped |
| add L2F attenuation | LEO · miniImageNet 5-way 5-shot | accuracy: 77.12 \pm 0.07 % -> 78.13 \pm 0.15% | Helped |
| add L2F attenuation | MAML · FC100 5-way 1-shot | accuracy: 35.98 \pm 0.48% -> 39.46 \pm 0.49% | Helped |
| add L2F attenuation | MAML · CIFAR-FS 5-way 5-shot | accuracy: 70.16 \pm 0.46% -> 73.94 \pm 0.44% | Helped |
| use task-independent layer-wise attenuation rather than no attenuation (ablation) | MAML · miniImageNet 5-way 5-shot, 4-layer CNN | accuracy: 63.94\pm0.48% -> 68.49 \pm 0.41% | Helped |
| replace layer-wise attenuation with network-wise attenuation (ablation) | MAML+L2F · miniImageNet 5-way 5-shot, 4-layer CNN | accuracy: 68.49 \pm 0.41% -> 67.84 \pm 0.46% | Hurt |
| replace layer-wise attenuation with filter-wise attenuation (ablation) | MAML+L2F · miniImageNet 5-way 5-shot, 4-layer CNN | accuracy: 68.49 \pm 0.41% -> 65.35 \pm 0.48% | Hurt |
| replace layer-wise attenuation with parameter-wise attenuation (ablation) | MAML+L2F · miniImageNet 5-way 5-shot, 4-layer CNN | accuracy: 68.49 \pm 0.41% -> 64.7 \pm 0.43% | Hurt |
| remove task-conditioning from layer-wise attenuation (ablation) | MAML+L2F · miniImageNet 5-way 5-shot | accuracy: 68.49 \pm 0.41% -> 69.38 \pm 0.46% | Hurt |
| replace sigmoided attenuation with task-dependent affine transformation (ablation) | MAML+L2F · miniImageNet 5-way 5-shot | accuracy: 66.22 \pm 0.47% -> 69.38 \pm 0.46% | Hurt |
| remove the sigmoid constraint from task-dependent gamma (ablation) | MAML+L2F · miniImageNet 5-way 5-shot | accuracy: 67.56 \pm 0.47% -> 69.38 \pm 0.46% | Hurt |
Closest papers by meaning
- Addressing Catastrophic Forgetting in Few-Shot Problems (ICML 2021 · similarity 0.89)
- Data-Centric Meta-Learning for Robust Few-Shot Generalization (CVPR 2026 · similarity 0.89)
- Probabilistic Model-Agnostic Meta-Learning (NeurIPS 2018 · similarity 0.88)
- First-order ANIL provably learns representations despite overparametrisation (ICLR 2024 · similarity 0.87)
- Provable Meta-Learning with Low-Rank Adaptations (NeurIPS 2025 · similarity 0.86)
- Memory-Reduced Meta-Learning with Guaranteed Convergence (AAAI 2025 · 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.