Research map / Learning theory and algorithms / Prediction and decision losses
Learning Deep Embeddings with Histogram Loss
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 |
|---|---|---|---|
| Histogram loss instead of Binomial Deviance (C=10) | Binomial Deviance · CUHK03-labeled | Recall@K: 2.64 | Helped |
| Histogram loss instead of Binomial Deviance (C=10 and 25) | Binomial Deviance · Market-1501 | Recall@K | Helped |
| Binomial Deviance with C=2 instead of C=25 (ablation) | Binomial Deviance · CUB-200-2011 | Recall@1: 43.50 | Hurt |
| Binomial Deviance C=25 instead of C=10 (ablation) | Binomial Deviance · Market-1501 | Recall@K | Helped |
| Binomial Deviance C=25 vs Histogram loss | Histogram loss · CUB-200-2011 and Online Products | Recall@K | Hurt |
| vary histogram bin size 0.005, 0.01, 0.02, 0.04 (400, 200, 100, 50 bins) (ablation) | Histogram loss · CUB-200-2011 | Recall@K | No clear effect |
| batch size 256 vs 128 and 64 (ablation) | Histogram loss · CUHK03 | Recall@K | Helped |
| batch size 256 vs smaller (ablation) | Histogram loss · Market-1501 | Recall@K | Helped |
| Histogram loss vs LSSS triplet loss | LSSS · CUB-200-2011, Online Products, CUHK03, Market-1501 | Recall@K | Helped |
| Histogram loss vs semi-hard triplet loss | Triplet loss · CUHK03, Market-1501 | Recall@K | Helped |
| add non-zero margin to histogram loss (ablation) | Histogram loss · preliminary | Recall@K | No clear effect |
Closest papers by meaning
- Embedding Transfer With Label Relaxation for Improved Metric Learning (CVPR 2021 · similarity 0.85)
- OLÉ: Orthogonal Low-Rank Embedding - A Plug and Play Geometric Loss for Deep Learning (CVPR 2018 · similarity 0.85)
- More Information Supervised Probabilistic Deep Face Embedding Learning (ICML 2020 · similarity 0.85)
- On the Similarities of Embeddings in Contrastive Learning (ICML 2025 · similarity 0.84)
- Engineering the Neural Collapse Geometry of Supervised-Contrastive Loss (Student Abstract) (AAAI 2024 · similarity 0.83)
- Efficient Distribution Matching of Representations via Noise-Injected Deep InfoMax (ICLR 2025 · 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.