Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Deconstructed Generation-Based Zero-Shot Model
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 attribute augmentation to generator training (ablation) | DGZ · AWA2 and CUB | H: 71.4 -> 73.6; 69.0 -> 69.2 | Helped |
| Use revised classifier loss instead of vanilla cross-entropy (ablation) | DGZ · AWA2 and CUB | H: 55.1 -> 73.6; 64.0 -> 69.2 | Helped |
| Add the classifier weight mapping net (ablation) | DGZ · AWA2 and CUB | H: 70.9 -> 73.6; 63.6 -> 69.2 | Helped |
| Keep classifier revision and mapping net rather than removing both (ablation) | DGZ · AWA2 and CUB | H: 50.0 -> 73.6; 54.7 -> 69.2 | Helped |
| Replace generated distribution with statistical-covariance Gaussian (ablation) | DGZ · AWA2 and CUB | H: 72.4 -> 73.6; 68.0 -> 69.2 | Hurt |
| Use attribute-to-class-center mapping with statistical-covariance Gaussian (ablation) | DGZ · AWA2 and CUB | H: 71.5 -> 73.6; 68.0 -> 69.2 | Hurt |
| Replace the generative model with a mapping net for class centers (ablation) | DGZ · AWA2, CUB, SUN, and APY | H: 73.6 -> 71.5; 69.2 -> 68.0; 43.5 -> 42.8; 47.7 -> 47.6 | Mixed |
| Add L2 regularization to the generator | WGAN on AWA2 · AWA2 | H: 63.1 -> 65.2 | Helped |
| Apply Fast Gradient Method to generator attributes | WGAN on AWA2 · AWA2 | H: 63.1 -> 65.6 | Helped |
| Augment generator attributes with Gaussian noise | WGAN on AWA2 · AWA2 | H: 63.1 -> 65.9 | Helped |
| Plug DGZ classifier learning into CVAE | VAE · AWA2, CUB, SUN, and APY | H: 61.2 -> 71.9; 60.7 -> 68.7; 39.6 -> 43.2; 40.6 -> 47.7 | Helped |
| Replace the mapping net with CN prototype learning | DGZ · AWA2 and CUB | H: 73.6 -> 73.1; 69.2 -> 66.4 | Hurt |
Closest papers by meaning
- Data Distribution Distilled Generative Model for Generalized Zero-Shot Recognition (AAAI 2024 · similarity 0.90)
- Meta-Learning for Generalized Zero-Shot Learning (AAAI 2020 · similarity 0.89)
- GenZSL: Generative Zero-Shot Learning Via Inductive Variational Autoencoder (ICML 2025 · similarity 0.89)
- Improving Generalized Zero-Shot Learning by Exploring the Diverse Semantics from External Class Names (CVPR 2024 · similarity 0.87)
- Data-Free Generalized Zero-Shot Learning (AAAI 2024 · similarity 0.87)
- Class Distribution Shifts in Zero-Shot Learning: Learning Robust Representations (NeurIPS 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.