Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Learning Feature-to-Feature Translator by Alternating Back-Propagation for Generative Zero-Shot Learning
Yizhe Zhu, Jianwen Xie, Bingchen Liu, Ahmed Elgammal
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 |
|---|---|---|---|
| replace GAN/VAE auxiliary networks with a single conditional generator trained by alternating back-propagation | generative ZSL · AwA2 (ZSL) | top-1 accuracy: 67.3 -> 70.4 | Improved |
| use alternating back-propagation generator instead of strongest generative baseline | generative GZSL · AwA2 (GZSL) | harmonic mean H: 58.6 -> 62.6 | Improved |
| alternating back-propagation generator vs MCGZSL on CUB | generative GZSL · CUB (GZSL) | harmonic mean H: 52.3 -> 50.6 | Hurt |
| drop discriminator/encoder auxiliary networks, keeping one generator | feature translator · CUB | parameters: 20.62M -> 9.71M | Improved |
| train with alternating back-propagation instead of adversarial training | feature translator · CUB/AwA1 | convergence speed | Improved |
| generate only one synthetic feature per unseen class | feature translator vs FGZSL · CUB | top-1 accuracy: 29.1 -> 50.1 | Improved |
| increase synthetic features per unseen class up to 300 (ablation) | feature translator · CUB/AwA1 | top-1 accuracy | Improved |
| learn from incomplete visual features with ours vs GAZSL (GTA features) | feature translator · CUB (GTA features) | top-1 accuracy: 74.1 -> 76.7 | Improved |
| raise missing ratio of DET features to 90% (ablation) | feature translator · CUB (DET features) | top-1 accuracy: 75.2 -> 51.6 | Hurt |
| increase number of seen classes in training (ablation) | generative models · SUN (GZSL) | unseen accuracy | Improved |
Closest papers by meaning
- Boosting Generative Zero-Shot Learning by Synthesizing Diverse Features with Attribute Augmentation (AAAI 2022 · similarity 0.91)
- GenZSL: Generative Zero-Shot Learning Via Inductive Variational Autoencoder (ICML 2025 · similarity 0.90)
- Zero-Shot Learning from Adversarial Feature Residual to Compact Visual Feature (AAAI 2020 · similarity 0.90)
- ZeroDiff: Solidified Visual-semantic Correlation in Zero-Shot Learning (ICLR 2025 · similarity 0.89)
- Visual-Augmented Dynamic Semantic Prototype for Generative Zero-Shot Learning (CVPR 2024 · similarity 0.88)
- Data-Free Generalized Zero-Shot Learning (AAAI 2024 · similarity 0.88)
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.
