Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Adaptive Distribution Calibration for Few-Shot Learning with Hierarchical Optimal Transport
Dan dan Guo, Long Tian, He Zhao, Mingyuan Zhou, Hongyuan Zha
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 hierarchical OT adaptive weights instead of top-k Euclidean selection | Free-Lunch · miniImageNet 5-way 1-shot | accuracy: 68.57 -> 69.04 | Improved |
| add hierarchical OT adaptive weights instead of top-k Euclidean selection | Free-Lunch · CUB 5-way 1-shot | accuracy: 79.56 -> 81.23 | Improved |
| add hierarchical OT adaptive weights | Free-Lunch · tieredImageNet 5-way 5-shot (ResNet12) | accuracy: 85.56 -> 87.46 | Improved |
| add low-level OT adaptive cost instead of cosine cost to class mean (ablation) | H-OT · miniImageNet 5-way 1-shot | accuracy: 68.82 -> 69.04 | Improved |
| add low-level OT adaptive cost instead of cosine cost to class mean (ablation) | H-OT · CUB 5-way 1-shot | accuracy: 81.01 -> 81.23 | Improved |
| add low-level OT adaptive cost instead of Euclidean cost to class mean (ablation) | H-OT · tieredImageNet 5-way 1-shot | accuracy: 75.31 -> 75.91 | Improved |
| use sample-weighted class mean in cost function (ablation) | H-OT · miniImageNet 5-way 1-shot | accuracy: 68.82 -> 68.78 | No clear effect |
| use cosine instead of Euclidean cost with high-level OT only (ablation) | H-OT · CUB 5-way 1-shot | accuracy: 80.29 -> 81.01 | Improved |
| use high-level OT with fixed cost instead of Free-Lunch top-k (ablation) | H-OT · miniImageNet 5-way 1-shot | accuracy: 68.57 -> 68.79 | Improved |
| swap WRN28+RTloss for weaker backbones (Conv4, ResNet18) | H-OT vs Free-Lunch · miniImageNet 5-way 1-shot | accuracy: 61.50 -> 62.61 (ResNet18) | Improved |
| cross-domain base from CIFAR-FS to novel CUB | H-OT · CIFAR-FS to CUB | accuracy | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 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 1 shared result
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
- Bridging the Gap Between Few-Shot and Many-Shot Learning via Distribution Calibration (TPAMI 2021 · similarity 0.89)
- Free Lunch for Few-shot Learning: Distribution Calibration (ICLR 2021 · similarity 0.89)
- Discriminative Sample-Guided and Parameter-Efficient Feature Space Adaptation for Cross-Domain Few-Shot Learning (CVPR 2024 · similarity 0.85)
- Provably Improving Generalization of Few-shot models with Synthetic Data (ICML 2025 · similarity 0.84)
- Dual-Level Curriculum Meta-Learning for Noisy Few-Shot Learning Tasks (AAAI 2024 · similarity 0.84)
- Few-shot Adaptation to Distribution Shifts By Mixing Source and Target Embeddings (ICML 2024 · similarity 0.84)
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.
