Research map / Label-efficient and robust learning / Recognition and long-tail
A unifying mutual information view of metric learning: cross-entropy vs. pairwise losses
Malik Boudiaf, Jérôme Rony, Imtiaz Masud Ziko, Eric Granger, Marco Pedersoli, Pablo Piantanida, Ismail Ben Ayed
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 |
|---|---|---|---|
| evaluate with cosine distance instead of Euclidean distance on unnormalized features | Cross-Entropy (ResNet-50) · CUB-200-2011 | Recall@1: 67.6 -> 69.2 | Improved |
| evaluate with cosine distance instead of Euclidean distance on unnormalized features | Cross-Entropy (ResNet-50) · Cars-196 | Recall@1: 89.1 -> 89.3 | No clear effect |
| evaluate with cosine distance instead of Euclidean distance on unnormalized features | Cross-Entropy (ResNet-50) · Stanford Online Products | Recall@1: 80.8 -> 81.1 | No clear effect |
| evaluate with cosine distance instead of Euclidean distance on unnormalized features | Cross-Entropy (ResNet-50) · In-Shop Clothes Retrieval | Recall@1: 90.6 -> 90.6 | No clear effect |
| add l2-normalization of features during training (ablation) | Cross-Entropy (ResNet-50) · four DML benchmarks | Recall | Hurt |
| add batch normalization (no scale or bias) on top of the feature extractor (ablation) | Cross-Entropy (ResNet-50) · Cars, SOP, In-Shop | Recall | Improved |
| add batch normalization (no scale or bias) on top of the feature extractor (ablation) | Cross-Entropy (ResNet-50) · CUB-200-2011 | Recall | Hurt |
| add random jittering of brightness, contrast and saturation (ablation) | Cross-Entropy (ResNet-50) · CUB and Cars | Recall | Improved |
| tune the learning rate per dataset rather than sharing one (ablation) | Cross-Entropy (ResNet-50) · four DML benchmarks | Recall | Improved |
| train with simplified pairwise cross-entropy (SPCE) instead of cross-entropy | small CNN · MNIST | validation accuracy | No clear effect |
Closest papers by meaning
- A Simple and Effective Framework for Pairwise Deep Metric Learning (ECCV 2020 · similarity 0.88)
- Multi-Similarity Loss With General Pair Weighting for Deep Metric Learning (CVPR 2019 · similarity 0.88)
- Sharing Matters for Generalization in Deep Metric Learning (TPAMI 2020 · similarity 0.87)
- Generative Cross-Entropy: A Strictly Proper Loss for Data-Efficient Classification (NeurIPS 2026 · similarity 0.87)
- Engineering the Neural Collapse Geometry of Supervised-Contrastive Loss (Student Abstract) (AAAI 2024 · similarity 0.85)
- Mean Field Theory in Deep Metric Learning (ICLR 2024 · 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.
