Research map / Label-efficient and robust learning / Recognition and long-tail
Unsupervised Metric Learning with Synthetic Examples
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 |
|---|---|---|---|
| learn metric with SUML instead of NPE/LPP manifold learning | kNN classification · AT&T Faces k=1 | accuracy (%): 84.92 -> 90.08 | Helped |
| unsupervised SUML instead of supervised DML (best baseline) | kNN classification · AT&T Faces k=1 | accuracy (%): 93.33 -> 90.08 | Hurt |
| unsupervised SUML instead of MOM unsupervised baseline | zero-shot clustering · JHMDB zero-shot | NMI: 65.21 -> 70.58 | Helped |
| SUML instead of classical supervised DML (DRIFT) | clustering with noise · AwA2+CUB noise | NMI: 80.19 -> 83.47 | Helped |
| sSUML instead of MOM | fine-grained retrieval · CUB 200 | Recall@1 (%): 45.3 -> 43.5 | Hurt |
| sSUML instead of NCE | fine-grained retrieval · Cars 196 | Recall@1 (%): 37.5 -> 42.0 | Helped |
| train sSUML from random initial embedding (ablation) | sSUML · Stanford Online Products | Recall@1 (%): 25.4 -> 47.8 | Helped |
| set orthogonality regularizer weight lambda too low (ablation) | SUML · JHMDB zero-shot | NMI and F | Hurt |
| vary angle alpha (ablation) | SUML · JHMDB zero-shot | NMI and F | No clear effect |
| stochastic sSUML versus supervised AML-stoch | large-scale metric learning · ImageNet | NMI: 58.1 -> 56.6 | Hurt |
Closest papers by meaning
- Learning Low-Dimensional Metrics (NeurIPS 2017 · similarity 0.90)
- Self-Taught Metric Learning Without Labels (CVPR 2022 · similarity 0.89)
- Semi-Supervised Metric Learning: A Deep Resurrection (AAAI 2021 · similarity 0.88)
- Large-Scale Distance Metric Learning With Uncertainty (CVPR 2018 · similarity 0.88)
- Provably Robust Metric Learning (NeurIPS 2020 · similarity 0.88)
- Potential Field Based Deep Metric Learning (CVPR 2025 · similarity 0.86)
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.