Research map / Label-efficient and robust learning / Domain adaptation and generalization
Diversity-enhancing Generative Network for Few-shot Hypothesis Adaptation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 HSIC diversity regularization to the generator (ablation) | DEG-Net · M→U, 7 target examples/class | accuracy (%): 91.2±0.5 -> 92.1±0.6 | Helped |
| add HSIC diversity regularization to the generator (ablation) | DEG-Net · M→U, 1 target example/class | accuracy (%): 87.2±1.9 -> 87.3±0.9 | No clear effect |
| replace separate class-specific generators with a weight-shared generator (ablation) | DEG-Net · M→U, 7 target examples/class | accuracy (%): 90.4 ±1.2 -> 92.1±0.6 | Helped |
| use DEG-Net rather than TOHAN | few-shot hypothesis adaptation · M→U, 7 target examples/class | accuracy (%): 89.7±0.5 -> 92.1±0.6 | Helped |
| use DEG-Net rather than TOHAN | few-shot hypothesis adaptation · M→S, 7 target examples/class | accuracy (%): 33.2±0.8 -> 34.9±1.6 | Helped |
| use DEG-Net rather than TOHAN | few-shot hypothesis adaptation · M→U, 1 target example/class | accuracy (%): 76.0±1.9 -> 83.1±0.9 | Helped |
| use DEG-Net rather than TOHAN | few-shot hypothesis adaptation · U→M, 1 target example/class | accuracy (%): 84.0±0.5 -> 82.2±0.7 | Hurt |
| use DEG-Net rather than TOHAN | few-shot hypothesis adaptation · VisDA-C, 8 target examples/class | accuracy (%): 64.75 -> 65.36 | Helped |
| use a shared conditional generator instead of separate class-specific generators (ablation) | DEG-Net · M→U | training time per epoch: 1.35s -> 0.93s | Helped |
| use DEG-Net rather than TOHAN to generate unlabeled data | generative FHA · M→S | HSIC diversity measure: 0.0027 -> 0.0019 | Helped |
Closest papers by meaning
- Diversity-Driven Synthesis: Enhancing Dataset Distillation through Directed Weight Adjustment (NeurIPS 2024 · similarity 0.86)
- Revisiting Source-Free Domain Adaptation: Insights into Representativeness, Generalization, and Variety (CVPR 2025 · similarity 0.84)
- Rethinking Few-Shot Adaptation of Vision-Language Models in Two Stages (CVPR 2025 · similarity 0.84)
- Improving Generalized Zero-Shot Learning by Exploring the Diverse Semantics from External Class Names (CVPR 2024 · similarity 0.84)
- Leveraging Normalization Layer in Adapters with Progressive Learning and Adaptive Distillation for Cross-Domain Few-Shot Learning (AAAI 2024 · similarity 0.84)
- Simple Semantic-Aided Few-Shot Learning (CVPR 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.