Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Learning Classifiers for Target Domain with Limited or No Labels
Pengkai Zhu, Hanxiao Wang, Venkatesh Saligrama
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 |
|---|---|---|---|
| learn encoder jointly on source and unlabeled target data with pseudo-label cross-entropy (joint pi) vs source-only (ablation) | LDVA · SVHN->MNIST | accuracy (%): 82.4 -> 95.2 | Improved |
| joint pi vs source pi (ablation) | LDVA · MNIST->USPS | accuracy (%): 94.8 -> 98.8 | Improved |
| joint pi vs source pi (ablation) | LDVA · USPS->MNIST | accuracy (%): 96.1 -> 96.8 | Improved |
| add calibrated stacking (CS) to LDVA (ablation) | LDVA · CUB GZSL | harmonic mean H (%): 48.4 -> 66.0 | Improved |
| add calibrated stacking (CS) (ablation) | LDVA · AWA2 GZSL | harmonic mean H (%): 57.2 -> 67.3 | Improved |
| add calibrated stacking (CS) (ablation) | LDVA · aPY GZSL | harmonic mean H (%): 36.6 -> 51.2 | Improved |
| add calibrated stacking (CS) lowers seen-class accuracy to balance (ablation) | LDVA · CUB GZSL | seen accuracy tr (%): 87.5 -> 74.6 | Hurt |
| add calibrated stacking raises unseen accuracy (ablation) | LDVA · CUB GZSL | unseen accuracy ts (%): 33.4 -> 59.2 | Improved |
| LDVA+CS vs second-best LSD | LDVA + CS · CUB GZSL | harmonic mean H (%): 56.1 -> 66.0 | Improved |
| LDVA+CS vs second-best CADA-VAE | LDVA + CS · AWA2 GZSL | harmonic mean H (%): 63.9 -> 67.3 | Improved |
| LDVA nearest-neighbour classifier without meta-training vs EA-FSL | LDVA · miniImageNet 5-way 5-shot | accuracy (%): 78.4 -> 78.7 | Improved |
| LDVA vs LEO | LDVA · miniImageNet 5-way 1-shot | accuracy (%): 61.7 -> 61.7 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- TADAM — TADAM: Task dependent adaptive metric for improved few-shot learning (NeurIPS 2018) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Rethinking Zero-Shot Learning: A Conditional Visual Classification Perspective (ICCV 2019 · similarity 0.90)
- Semantic Autoencoder for Zero-Shot Learning (CVPR 2017 · similarity 0.90)
- Classifier and Exemplar Synthesis for Zero-Shot Learning (IJCV 2019 · similarity 0.90)
- Domain-Invariant Projection Learning for Zero-Shot Recognition (NeurIPS 2018 · similarity 0.89)
- An Empirical Study and Analysis of Generalized Zero-Shot Learning for Object Recognition in the Wild (ECCV 2016 · similarity 0.89)
- Data-Free Generalized Zero-Shot Learning (AAAI 2024 · similarity 0.89)
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.
