Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Meta Navigator: Search for a Good Adaptation Policy for Few-Shot Learning
Chi Zhang, Henghui Ding, Guosheng Lin, Ruibo Li, Changhu Wang, Chunhua Shen
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 |
|---|---|---|---|
| search adaptation policies per stage instead of randomly sampling policies (ablation) | random search · miniImageNet 5-way 5-shot | accuracy: 77.97 -> 82.66 | Improved |
| search adaptation policies per stage instead of randomly sampling policies (ablation) | random search · miniImageNet 5-way 1-shot | accuracy: 64.10 -> 65.91 | Improved |
| search adaptation policies instead of random sampling (cross-domain) (ablation) | random search · miniImageNet to CUB 10-shot | accuracy: 71.64 -> 81.05 | Improved |
| search policies instead of fixed Baseline++ policy (cross-domain) (ablation) | Baseline++ · miniImageNet to CUB 10-shot | accuracy: 76.13 -> 81.05 | Improved |
| use searched policy instead of plain fine-tuning | finetune · miniImageNet 5-way 1-shot | accuracy: 44.81 -> 65.91 | Improved |
| use searched policy instead of MAML policy | MAML · miniImageNet 5-way 5-shot | accuracy: 75.80 -> 82.66 | Improved |
| use searched policy instead of ProtoNets policy | ProtoNets · miniImageNet 5-way 10-shot | accuracy: 82.66 -> 85.46 | Improved |
| search policy instead of fixed Baseline++ on cross-domain 5-shot | Baseline++ · miniImageNet to CUB 5-shot | accuracy: 70.01 -> 72.43 | Improved |
| add multi-crop testing (ablation) | Meta Navigator · tieredImageNet 5-way 1-shot | accuracy: 73.52 -> 74.58 | Improved |
| add multi-crop testing (ablation) | Meta Navigator · miniImageNet 5-way 1-shot | accuracy: 65.91 -> 67.14 | Improved |
| add multi-crop testing (ablation) | Meta Navigator · FC100 5-way 5-shot | accuracy: 59.93 -> 61.33 | Improved |
| compare with DeepEMD on FC100 5-shot | Meta Navigator · FC100 5-way 5-shot | accuracy: 63.22 vs 59.93 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- DeepEMD — DeepEMD: Few-Shot Image Classification With Differentiable Earth Mover's Distance and Structured Classifiers (CVPR 2020) · better on 1 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Meta-Baseline: Exploring Simple Meta-Learning for Few-Shot Learning (ICCV 2021 · similarity 0.92)
- Few-Shot Learning With a Strong Teacher (TPAMI 2022 · similarity 0.91)
- Meta-Dataset: A Dataset of Datasets for Learning to Learn from Few Examples (ICLR 2020 · similarity 0.91)
- Learning to Learn Task-Adaptive Hyperparameters for Few-Shot Learning (TPAMI 2023 · similarity 0.91)
- A Theoretical Analysis of the Number of Shots in Few-Shot Learning (ICLR 2020 · similarity 0.90)
- Meta-Learning With Task-Adaptive Loss Function for Few-Shot Learning (ICCV 2021 · similarity 0.90)
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.
