Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Adversarial Feature Augmentation for Cross-Domain Few-Shot Classification
Yanxu Hu, Andy J. Ma
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 AFA module to GNN | GNN · CUB 5-shot | accuracy (%): 62.87 -> 68.25 | Improved |
| add AFA module to meta-learning baselines | MN, GNN, TPN · eight target datasets, 1-shot | accuracy (%): 3.45% | Improved |
| add AFA module to meta-learning baselines | MN, GNN, TPN · eight target datasets, 5-shot | accuracy (%): 4.25% | Improved |
| use AFA instead of ATA | meta-learning baselines · eight target datasets | accuracy (%): about 1% | Improved |
| add domain discriminator to AFA (ablation) | MN+AFA · eight target datasets, 1-shot | accuracy (%): 2.48% | Improved |
| add domain discriminator to AFA (ablation) | MN+AFA · CUB 5-shot | accuracy (%): 58.06 -> 59.46 | Improved |
| add gram-matrix loss to AFA (ablation) | MN+AFA · eight target datasets | accuracy (%): 2.57% | Improved |
| add gram-matrix loss to AFA (ablation) | MN+AFA · CUB 5-shot | accuracy (%): 56.44 -> 59.46 | Improved |
| replace linear mean/variance perturbation with non-linear convolution (ablation) | MN+AFA · CUB 5-shot | accuracy (%): 59.46 -> 50.17 | Hurt |
| replace linear perturbation with non-linear convolution (ablation) | MN+AFA · CUB 1-shot | accuracy (%): 41.02 -> 34.42 | Hurt |
| AFA trained without domain discriminator (classification loss only) (ablation) | MN · CropDiseases | accuracy (%): declines vs baseline MN | Hurt |
| use meta-learning with AFA instead of pre-train and fine-tuning | GNN · CUB 5-shot | accuracy (%): 63.76 -> 70.33 | Improved |
Closest papers by meaning
- Few-Shot Adversarial Domain Adaptation (NeurIPS 2017 · similarity 0.90)
- Discriminative Sample-Guided and Parameter-Efficient Feature Space Adaptation for Cross-Domain Few-Shot Learning (CVPR 2024 · similarity 0.88)
- Cross-Domain Few-Shot Classification via Learned Feature-Wise Transformation (ICLR 2020 · similarity 0.88)
- Selecting Relevant Features from a Multi-domain Representation for Few-shot Classification (ECCV 2020 · similarity 0.88)
- MetaGAN: An Adversarial Approach to Few-Shot Learning (NeurIPS 2018 · similarity 0.88)
- Pushing the Limit of Fine-Tuning for Few-Shot Learning: Where Feature Reusing Meets Cross-Scale Attention (AAAI 2024 · 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.
