Research map / Learning theory and algorithms / Kernels and regression theory
Posterior Refinement on Metric Matrix Improves Generalization Bound in Metric Learning
Mingda Wang, Canqian Yang, Yi Xu
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 trained metric matrix with an identity matrix | ArcFace · CUB-200-2011 | Precision@1: 60.12 -> 65.06 | Improved |
| replace trained metric matrix with an identity matrix | ProxyNCA · CUB-200-2011 | Precision@1: 56.99 -> 63.64 | Improved |
| replace trained metric matrix with an identity matrix | Triplet · CUB-200-2011 | Precision@1: 54.95 -> 61.52 | Improved |
| replace trained metric matrix with an identity matrix | ArcFace · Cars196 | Precision@1: 72.88 -> 78.79 | Improved |
| replace trained metric matrix with an identity matrix | ProxyNCA · Cars196 | Precision@1: 73.91 -> 80.10 | Improved |
| replace trained metric matrix with an identity matrix | Triplet · Cars196 | Precision@1: 64.92 -> 74.86 | Improved |
| replace trained metric matrix with an identity matrix | Triplet · SOP | Precision@1: 67.96 -> 70.49 | Improved |
| replace trained metric matrix with an identity matrix | ArcFace · SOP | Precision@1: 71.98 -> 72.28 | No clear effect |
| replace trained metric matrix with an identity matrix | ArcFace · SOP | MAP: 42.23 -> 42.18 | No clear effect |
| use best diagonal or random restraint instead of identity | ArcFace · CUB-200-2011 | Precision@1: 65.06 -> 64.90 | No clear effect |
| increase diagonal restraint degree r (reduce trace) (ablation) | ArcFace, ProxyNCA, Triplet · CUB-200-2011 | MAP, Precision@1, RP | Improved |
| increase random restraint degree r (reduce Frobenius norm) (ablation) | ArcFace, ProxyNCA, Triplet · CUB-200-2011 | MAP, Precision@1, RP | Improved |
Closest papers by meaning
- Revisiting Training Strategies and Generalization Performance in Deep Metric Learning (ICML 2020 · similarity 0.89)
- Characterizing Generalization under Out-Of-Distribution Shifts in Deep Metric Learning (NeurIPS 2021 · similarity 0.87)
- Learning with Diversity: Self-Expanded Equalization for Better Generalized Deep Metric Learning (ICCV 2023 · similarity 0.87)
- Generalization Bound of Gradient Descent for Non-Convex Metric Learning (NeurIPS 2020 · similarity 0.87)
- Deep Variational Metric Learning (ECCV 2018 · similarity 0.87)
- Deep Disentangled Metric Learning (AAAI 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.
