Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Hardness Sampling for Self-Training Based Transductive Zero-Shot Learning
Check 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 |
|---|---|---|---|
| Select diverse hard-class pseudo-labeled samples rather than random samples (ablation) | STHS-S2V · AWA2, CUB, SUN | ACC: 85.0 -> 91.4; 66.0 -> 71.2; 66.7 -> 70.9 | Helped |
| Select diverse hard-class pseudo-labeled samples rather than random samples (ablation) | STHS-WGAN · AWA2, CUB, SUN | ACC: 87.5 -> 94.9; 73.1 -> 77.4; 63.1 -> 67.5 | Helped |
| Use prior-normalized class frequency instead of raw class frequency (ablation) | STHS-S2V · AWA2 | ACC: 88.9 -> 91.4 | Helped |
| Use prior-normalized class frequency instead of raw class frequency (ablation) | STHS-WGAN · AWA2 | ACC: 89.8 -> 94.9 | Helped |
| Use prior-normalized class frequency instead of raw class frequency (ablation) | STHS-S2V · CUB and SUN | ACC: 71.2 -> 70.0; 70.9 -> 70.0 | Hurt |
| Use prior-normalized class frequency instead of raw class frequency (ablation) | STHS-WGAN · CUB and SUN | ACC: 77.4 -> 74.5; 67.5 -> 65.5 | Hurt |
| Add samples from hard classes rather than easy classes (ablation) | DEM and f-CLSWGAN · AWA2 | ACC | Helped |
| Use hard-class pseudo labels rather than easy-class pseudo labels (ablation) | DEM and f-CLSWGAN · AWA2 | Precision | Helped |
| Sample from Top-4 hard classes rather than Top-1 or Top-2 hard classes (ablation) | DEM and f-CLSWGAN · AWA2 | ACC | Helped |
| Use hardness sampling rather than random sampling (ablation) | STHS-GZSL · AWA2 strict T-GZSL | H: 78.7 -> 83.2 | Helped |
| Use hardness sampling rather than random sampling (ablation) | STHS-GZSL · CUB strict T-GZSL | H: 63.9 -> 66.2 | Helped |
| Use hardness sampling rather than random sampling (ablation) | STHS-GZSL · SUN strict T-GZSL | H: 38.6 -> 40.2 | Helped |
Closest papers by meaning
- Adaptive Confidence Smoothing for Generalized Zero-Shot Learning (CVPR 2019 · similarity 0.89)
- Meta-Learning for Generalized Zero-Shot Learning (AAAI 2020 · similarity 0.89)
- Bi-Directional Distribution Alignment for Transductive Zero-Shot Learning (CVPR 2023 · similarity 0.89)
- Dual-View Ranking with Hardness Assessment for Zero-Shot Learning (AAAI 2019 · similarity 0.88)
- Data Distribution Distilled Generative Model for Generalized Zero-Shot Recognition (AAAI 2024 · similarity 0.87)
- Improving Generalized Zero-Shot Learning by Exploring the Diverse Semantics from External Class Names (CVPR 2024 · similarity 0.86)
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.