Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Latent Embeddings for Zero-Shot Classification
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 |
|---|---|---|---|
| Replace SJE's single bilinear compatibility map with LatEm's latent piecewise-linear maps using w2v embeddings | Zero-shot classification · CUB | average per-class Top-1 accuracy: 28.4 -> 31.8 | Helped |
| Replace SJE's single bilinear compatibility map with LatEm's latent piecewise-linear maps using glo embeddings | Zero-shot classification · CUB | average per-class Top-1 accuracy: 24.2 -> 32.5 | Helped |
| Replace SJE's single bilinear compatibility map with LatEm's latent piecewise-linear maps using hie embeddings | Zero-shot classification · CUB | average per-class Top-1 accuracy: 20.6 -> 24.2 | Helped |
| Replace SJE's single bilinear compatibility map with LatEm's latent piecewise-linear maps using att embeddings | Zero-shot classification · CUB | average per-class Top-1 accuracy: 50.1 -> 45.5 | Hurt |
| Replace SJE's single bilinear compatibility map with LatEm's latent piecewise-linear maps using att embeddings | Zero-shot classification · AWA | average per-class Top-1 accuracy: 66.7 -> 71.9 | Helped |
| Replace SJE's single bilinear compatibility map with LatEm's latent piecewise-linear maps using w2v embeddings | Zero-shot classification · AWA | average per-class Top-1 accuracy: 51.2 -> 61.1 | Helped |
| Replace SJE's single bilinear compatibility map with LatEm's latent piecewise-linear maps using glo embeddings | Zero-shot classification · AWA | average per-class Top-1 accuracy: 58.8 -> 62.9 | Helped |
| Replace SJE's single bilinear compatibility map with LatEm's latent piecewise-linear maps using hie embeddings | Zero-shot classification · AWA | average per-class Top-1 accuracy: 51.2 -> 57.5 | Helped |
| Replace SJE's single bilinear compatibility map with LatEm's latent piecewise-linear maps using unsupervised embeddings | Zero-shot classification · Dogs | average per-class Top-1 accuracy: 19.6 -> 22.6; 17.8 -> 20.9; 24.3 -> 25.2 | Helped |
| Combine unsupervised class embeddings (w2v, glo, hie) rather than use individual embeddings | LatEm · CUB | average per-class Top-1 accuracy: 29.9 -> 34.9 | Helped |
| Combine unsupervised class embeddings (w2v, glo, hie) rather than use individual embeddings | LatEm · AWA | average per-class Top-1 accuracy: 60.1 -> 66.2 | Helped |
| Combine unsupervised class embeddings (w2v, glo, hie) rather than use individual embeddings | LatEm · Dogs | average per-class Top-1 accuracy: 35.1 -> 36.3 | Helped |
Closest papers by meaning
- Generalized Zero- and Few-Shot Learning via Aligned Variational Autoencoders (CVPR 2019 · similarity 0.91)
- Zero-Shot Learning via Joint Latent Similarity Embedding (CVPR 2016 · similarity 0.91)
- Hierarchical Disentanglement of Discriminative Latent Features for Zero-Shot Learning (CVPR 2019 · similarity 0.89)
- Adaptive and Generative Zero-Shot Learning (ICLR 2021 · similarity 0.88)
- Learning Aligned Cross-Modal Representation for Generalized Zero-Shot Classification (AAAI 2022 · similarity 0.88)
- TPR: Topology-Preserving Reservoirs for Generalized Zero-Shot Learning (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.