Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Meta Variance Transfer: Learning to Augment from the Others
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 |
|---|---|---|---|
| add meta variance transfer (MVT) to Matching network | Matching network (Conv4+) · miniImageNet 5-way 5-shot | accuracy: +2.79 (64.88 to 67.67) | Helped |
| add meta variance transfer (MVT) to Matching network | Matching network (Conv4+) · CUB 5-way 5-shot | accuracy: +3.35 (76.98 to 80.33) | Helped |
| add meta variance transfer (MVT) to Matching network | Matching network (Conv4+) · miniImageNet 5-way 10-shot | accuracy: +2.98 (68.85 to 71.83) | Helped |
| add meta variance transfer (MVT) to ResNet-10 backbone | ResNet-10 Matching network · CUB 5-way 5-shot | accuracy: +1.34 (84.01 to 85.35) | Helped |
| add meta variance transfer (MVT) to Matching network with Conv4 backbone | Matching network (Conv4) · miniImageNet 5-way 5-shot | accuracy: +0.61 (64.63 to 65.24) | Helped |
| add meta variance transfer (MVT) to Matching network | Matching network (Conv4+) on Omniglot and mini-VGG · mini-VGG 5-way 5-shot | accuracy: +1.18 (85.95 to 87.13) | Helped |
| naive variance transfer by adding one class's variations to another class mean without learning (ablation) | Matching network baseline · miniImageNet 5-way 5-shot | accuracy: -32.13 (64.88 to 32.76) | Hurt |
| add manifold regularization loss alone (ablation) | Matching network baseline · miniImageNet 5-way 5-shot | accuracy: -0.51 (64.88 to 64.37) | No clear effect |
| add manifold regularization to meta variance transfer (ablation) | Meta variance transfer · miniImageNet 5-way 5-shot | accuracy: +0.99 (66.68 to 67.67) | Helped |
| add manifold regularization to meta variance transfer (ablation) | Meta variance transfer · CUB 5-way 5-shot | accuracy: +0.41 (79.92 to 80.33) | Helped |
| add MVT on top of conventional augmentations (random crop, flip, color jitter) (ablation) | ResNet-10+ Matching network · CUB 5-shot / miniVGG / miniCASIA 5-shot | accuracy: CUB +1.34; miniVGG +0.33; miniCASIA +1.65 (with aug) | Helped |
| add MVT to Matching network across 10-way and 20-way 5-shot | Matching network (Conv4+) · miniImageNet 10-way / 20-way 5-shot | accuracy: +2.80 (10-way); +2.72 (20-way) | Helped |
Closest papers by meaning
- Meta-Transfer Learning for Few-Shot Learning (CVPR 2019 · similarity 0.88)
- On sensitivity of meta-learning to support data (NeurIPS 2021 · similarity 0.87)
- A Unified Few-Shot Classification Benchmark to Compare Transfer and Meta Learning Approaches (NeurIPS 2021 · similarity 0.87)
- Unleashing the Power of Meta-tuning for Few-shot Generalization Through Sparse Interpolated Experts (ICML 2024 · similarity 0.85)
- Pushing the Limit of Fine-Tuning for Few-Shot Learning: Where Feature Reusing Meets Cross-Scale Attention (AAAI 2024 · similarity 0.85)
- Provable Meta-Learning with Low-Rank Adaptations (NeurIPS 2025 · similarity 0.85)
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.