Research map / Graph learning / Graph algorithms and clustering
Fewer is More: A Deep Graph Metric Learning Perspective Using Fewer Proxies
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 |
|---|---|---|---|
| Use ProxyGML rather than SoftTriple | Deep metric learning · CUB-200-2011 | Recall@1: 65.4 -> 66.6 | Helped |
| Use ProxyGML rather than SoftTriple | Deep metric learning · Cars196 | Recall@1: 84.5 -> 85.5 | Helped |
| Use ProxyGML rather than SoftTriple | Deep metric learning · Stanford Online Products | NMI: 92.0 -> 90.2 | Hurt |
| Add the positive mask to the base loss (ablation) | ProxyGML · Cars196 | Recall@1: 47.3 -> 83.3 | Helped |
| Add mask softmax to the base loss (ablation) | ProxyGML · Cars196 | Recall@1: 47.3 -> 66.1 | Helped |
| Add proxy regularization to the base loss (ablation) | ProxyGML · Cars196 | Recall@1: 47.3 -> 81.7 | Helped |
| Add mask softmax and proxy regularization, omitting the positive mask (ablation) | ProxyGML · Cars196 | Recall@1: 47.3 -> 82.6 | Helped |
| Add the positive mask and proxy regularization, omitting mask softmax (ablation) | ProxyGML · Cars196 | Recall@1: 47.3 -> 84.5 | Helped |
| Add positive mask and mask softmax, omitting proxy regularization (ablation) | ProxyGML · Cars196 | Recall@1: 47.3 -> 84.0 | Helped |
| Use all three modules instead of the positive-mask-plus-mask-softmax variant (ablation) | ProxyGML · Cars196 | Recall@1: 84.0 -> 85.5 | Helped |
| Increase proxy count beyond 12 (ablation) | ProxyGML · Cars196 | Recall@1 | Hurt |
| Use a larger subgraph ratio than 0.05 (ablation) | ProxyGML with N=12 · Cars196 | Recall@1 | Hurt |
Closest papers by meaning
- Deep Metric Learning with Graph Consistency (AAAI 2021 · similarity 0.88)
- Proxy Anchor Loss for Deep Metric Learning (CVPR 2020 · similarity 0.88)
- Non-Isotropy Regularization for Proxy-Based Deep Metric Learning (CVPR 2022 · similarity 0.87)
- Towards Improved Proxy-Based Deep Metric Learning via Data-Augmented Domain Adaptation (AAAI 2024 · similarity 0.86)
- How Low Can You Go? Searching for the Intrinsic Dimensionality of Complex Networks using Metric Node Embeddings (ICLR 2025 · similarity 0.86)
- Deep Disentangled Metric Learning (AAAI 2025 · similarity 0.85)
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.