Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
ProxyNCA++: Revisiting and Revitalizing Proxy Neighborhood Component Analysis
Eu Wern Teh, Terrance DeVries, Graham W. Taylor
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 |
|---|---|---|---|
| add low temperature scaling (ablation) | ProxyNCA++ · CUB200-2011 | Recall@1: 61.4 -> 72.2 | Improved |
| add Global Max Pooling instead of average pooling (ablation) | ProxyNCA++ · CUB200-2011 | Recall@1: 69.0 -> 72.2 | Improved |
| add layer normalization (ablation) | ProxyNCA++ · CUB200-2011 | Recall@1: 69.6 -> 72.2 | Improved |
| add class balanced sampling (ablation) | ProxyNCA++ · CUB200-2011 | Recall@1: 69.6 -> 72.2 | Improved |
| add fast moving proxies (ablation) | ProxyNCA++ · CUB200-2011 | Recall@1: 70.3 -> 72.2 | Improved |
| add proxy assignment probability (ablation) | ProxyNCA++ · CUB200-2011 | Recall@1: 71.1 -> 72.2 | Improved |
| add proxy assignment probability to plain ProxyNCA (ablation) | ProxyNCA · CUB200-2011 | Recall@1: 59.3 -> 59.0 | No clear effect |
| add low temperature scaling to plain ProxyNCA (ablation) | ProxyNCA · CUB200-2011 | Recall@1: 59.3 -> 62.9 | Improved |
| add Layer Norm / CBS / fast proxies without low temperature scaling (ablation) | ProxyNCA · CUB200-2011 | Recall@1: 59.3 -> 54.8; 59.3 -> 59.0; 59.3 -> 56.3 | Hurt |
| replace GAP with GMP (ablation) | Margin · CUB200-2011 | Recall@1: 63.3 -> 64.3 | Improved |
| replace GAP with GMP (ablation) | Triplet-Semihard · CUB200-2011 | Recall@1: 60.5 -> 61.6 | Improved |
| replace GAP with GMP (ablation) | HORDE (Contrastive Loss) · CUB200-2011 | Recall@1: 65.1 -> 63.1 | Hurt |
Closest papers by meaning
- No Fuss Distance Metric Learning Using Proxies (ICCV 2017 · similarity 0.87)
- Towards Improved Proxy-Based Deep Metric Learning via Data-Augmented Domain Adaptation (AAAI 2024 · similarity 0.86)
- Proxy-FDA: Proxy-based Feature Distribution Alignment for Fine-tuning Vision Foundation Models without Forgetting (ICML 2025 · similarity 0.84)
- Stationary Representations: Optimally Approximating Compatibility and Implications for Improved Model Replacements (CVPR 2024 · similarity 0.82)
- Imagine and Seek: Improving Composed Image Retrieval with an Imagined Proxy (CVPR 2025 · similarity 0.82)
- Unsupervised Cross-Domain Image Retrieval via Prototypical Optimal Transport (AAAI 2024 · similarity 0.82)
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.
