Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
BOIL: Towards Representation Change for Few-shot 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 |
|---|---|---|---|
| Use BOIL, updating the body while freezing the head in the inner loop, instead of ANIL | 4conv few-shot image classification · Cars, CUB, miniImageNet, and tieredImageNet 5-way 5-shot | accuracy: miniImageNet: 63.04 $\pm$ 0.42 -> 66.45 $\pm$ 0.37; tieredImageNet: 65.82 $\pm$ 0.12 -> 69.37 $\pm$ 0.12; Cars: 61.95 $\ | Helped |
| Freeze the head during inner-loop adaptation rather than give it learning rate 0.05 (ablation) | BOIL with 4conv network · miniImageNet and Cars 5-way 5-shot | accuracy: miniImageNet: 38.81 $\pm$ 0.21 -> 66.45 $\pm$ 0.37; Cars: 68.67 $\pm$ 0.21 -> 75.18 $\pm$ 0.21 | Helped |
| Disconnect the last skip connection (ablation) | BOIL with ResNet-12 · 5-way 5-shot, all reported meta-train/meta-test settings | accuracy: miniImageNet: 70.50 $\pm$ 0.28 -> 71.30 $\pm$ 0.28; tieredImageNet: 71.86 $\pm$ 0.21 -> 74.12 $\pm$ 0.30; Cars: 49.69 $\ | Mixed |
| Adapt the body by updating conv3 alone rather than conv1 alone (ablation) | BOIL-style single-layer adaptation with 4conv · miniImageNet 5-way 5-shot | accuracy: 54.74 -> 67.07 | Helped |
Closest papers by meaning
- Rapid Learning or Feature Reuse? Towards Understanding the Effectiveness of MAML (ICLR 2020 · similarity 0.91)
- MAML and ANIL Provably Learn Representations (ICML 2022 · similarity 0.90)
- How to Train Your MAML to Excel in Few-Shot Classification (ICLR 2022 · similarity 0.89)
- Repurposing Pretrained Models for Robust Out-of-domain Few-Shot Learning (ICLR 2021 · similarity 0.89)
- How to train your MAML (ICLR 2019 · similarity 0.89)
- Provable Meta-Learning with Low-Rank Adaptations (NeurIPS 2025 · similarity 0.87)
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.