Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Partner-Assisted Learning for Few-Shot Image Classification
Jiawei Ma, Hanchen Xie, Guangxing Han, Shih-Fu Chang, Aram Galstyan, Wael Abd-Almageed
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 partner-assisted alignment (SupCT partner -> CE main) over CE-only training (ablation) | ResNet-12 main encoder · miniImageNet 5-way 1-shot | accuracy (%): 63.76 -> 69.37 | Improved |
| add partner-assisted alignment over CE-only training (ablation) | ResNet-12 main encoder · miniImageNet 5-way 5-shot | accuracy (%): 81.17 -> 84.40 | Improved |
| use SupCT partner one-way instead of multi-task L_CE + L_SupCT (ablation) | PAL · miniImageNet 5-way 1-shot | accuracy (%): 67.53 -> 69.37 | Improved |
| use one-way SupCT partner instead of mutual learning (ablation) | PAL · miniImageNet 5-way 1-shot | accuracy (%): 65.21 -> 69.37 | Improved |
| use SupCT -> CE direction instead of CE -> SupCT (ablation) | PAL · miniImageNet 5-way 1-shot | accuracy (%): 66.54 -> 69.37 | Improved |
| train partner with SupCT instead of CE (ablation) | PAL · miniImageNet 5-way 1-shot | accuracy (%): 66.95 -> 69.37 | Improved |
| train partner with SupCT instead of unsupervised contrastive (ablation) | PAL · miniImageNet 5-way 1-shot | accuracy (%): 65.89 -> 69.37 | Improved |
| add feature-level alignment to logit-level alignment (ablation) | PAL · miniImageNet 5-way 1-shot | accuracy (%): 66.35 -> 69.37 | Improved |
| add logit-level alignment to feature-level alignment (ablation) | PAL · miniImageNet 5-way 1-shot | accuracy (%): 68.03 -> 69.37 | Improved |
| feature-level alignment alone over no alignment (ablation) | PAL · miniImageNet 5-way 1-shot | accuracy (%): 63.76 -> 68.03 | Improved |
| use L_KL instead of L_logit with feature alignment (ablation) | PAL · miniImageNet 5-way 1-shot | accuracy (%): 69.37 -> 68.75 | Hurt |
| use L_KL alone instead of L_logit alone (ablation) | PAL · miniImageNet 5-way 5-shot | accuracy (%): 81.32 -> 80.58 | Hurt |
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 4 of 4 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
- LibFewShot: A Comprehensive Library for Few-Shot Learning (TPAMI 2023 · similarity 0.89)
- CAD: Co-Adapting Discriminative Features for Improved Few-Shot Classification (CVPR 2022 · similarity 0.89)
- A Closer Look at Few-shot Classification Again (ICML 2023 · similarity 0.89)
- Rethinking Generalization in Few-Shot Classification (NeurIPS 2022 · similarity 0.88)
- Pushing the Limit of Fine-Tuning for Few-Shot Learning: Where Feature Reusing Meets Cross-Scale Attention (AAAI 2024 · similarity 0.88)
- Simple Semantic-Aided Few-Shot Learning (CVPR 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.
