Research map / Label-efficient and robust learning / Self-supervised and contrastive
Pareto Self-Supervised Training for Few-Shot Learning
Zhengyu Chen, Jixie Ge, Heshen Zhan, Siteng Huang, Donglin Wang
Check 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 |
|---|---|---|---|
| Use PSST with rotation prediction instead of the prior auxiliary-training setup | MetaOptNet · MiniImageNet 1-shot | accuracy: 62.93 ± 0.63 -> 64.05± 0.49 | Improved |
| Use PSST with rotation prediction instead of the prior auxiliary-training setup | MetaOptNet · MiniImageNet 5-shot | accuracy: 79.63 ± 0.47 -> 80.24 ± 0.45 | Improved |
| Use PSST with location prediction instead of BF3S auxiliary training | CC · MiniImageNet 1-shot | accuracy: 60.71 ± 0.46 -> 63.26 ± 0.44 | Improved |
| Use PSST with location prediction instead of BF3S auxiliary training | CC · MiniImageNet 5-shot | accuracy: 77.64 ± 0.34 -> 80.05 ± 0.34 | Improved |
| Use PSST with rotation prediction instead of BF3S auxiliary training | CC · MiniImageNet 1-shot | accuracy: 62.93 ± 0.45 -> 64.16 ± 0.44 | Improved |
| Use PSST with rotation prediction instead of BF3S auxiliary training | CC · MiniImageNet 5-shot | accuracy: 79.87 ± 0.33 -> 80.64 ± 0.32 | Improved |
| Use PSST with rotation prediction instead of SSFSL auxiliary training | ProtoNet · MiniImageNet 1-shot | accuracy: 57.92 ± 0.58 -> 59.28 ± 0.46 | Improved |
| Use PSST with jigsaw prediction instead of SSFSL auxiliary training | ProtoNet · MiniImageNet 5-shot | accuracy: 76.20 ± 0.53 -> 77.27 ± 0.43 | Improved |
| Add location prediction without PSST (ablation) | CC · MiniImageNet 1-shot | accuracy: 61.09 ± 0.44 -> 60.71 ± 0.46 | Hurt |
| Add location prediction without PSST (ablation) | CC · MiniImageNet 5-shot | accuracy: 78.43 ± 0.33 -> 77.64 ± 0.34 | Hurt |
| Use PSST instead of grid-search weights to obtain multiple trade-off solutions | BF3S · Training, 10 solutions | iterations: 249 -> 60 | Improved |
| Restrict Pareto exploration to the preferred region | Pareto exploration on MultiMNIST · MultiMNIST | iterations: 130 -> 78 | Improved |
Closest papers by meaning
- Self-Supervision Can Be a Good Few-Shot Learner (ECCV 2022 · similarity 0.89)
- Task-Level Self-Supervision for Cross-Domain Few-Shot Learning (AAAI 2022 · similarity 0.87)
- Boosting Few-Shot Visual Learning With Self-Supervision (ICCV 2019 · similarity 0.87)
- TAFSSL: Task-Adaptive Feature Sub-Space Learning for few-shot classification (ECCV 2020 · similarity 0.87)
- Learning to Self-Train for Semi-Supervised Few-Shot Classification (NeurIPS 2019 · similarity 0.87)
- Pareto Low-Rank Adapters: Efficient Multi-Task Learning with Preferences (ICLR 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.
