Research map / Learning theory and algorithms / Prediction and decision losses
One Loss for Quantization: Deep Hashing With Discrete Wasserstein Distributional Matching
Check 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 |
|---|---|---|---|
| Replace the original quantization objective with HSWD (ablation) | DBDH · CIFAR-10, 32 bits | mAP: 0.8223 -> 0.8593 | Helped |
| Replace the original quantization objective with SWD (ablation) | DBDH · CIFAR-10, 32 bits | mAP: 0.8223 -> 0.8587 | Helped |
| Replace the original quantization objective with HSWD (ablation) | DBDH · NUS-WIDE, 32 bits | mAP: 0.8552 -> 0.8633 | Helped |
| Replace the original quantization objective with SWD (ablation) | DBDH · COCO, 32 bits | mAP: 0.7826 -> 0.7996 | Helped |
| Replace the original quantization objective with HSWD (ablation) | HashNet · NUS-WIDE, 32 bits | P@1000: 0.8158 -> 0.8078 | Hurt |
| Replace the original quantization objective with SWD (ablation) | GreedyHash · NUS-WIDE, 32 bits | P@1000: 0.8009 -> 0.7973 | Hurt |
| Replace the original quantization objective with HSWD (ablation) | GreedyHash · NUS-WIDE, 32 bits | P@1000: 0.8009 -> 0.7931 | Hurt |
| Replace the original quantization objective with HSWD (ablation) | DSDH · COCO, 16 bits | mAP: 0.7331 -> 0.7330 | Hurt |
| Use HSWD instead of SWD for the quantization objective (ablation) | DBDH · CIFAR-10, 32 bits | mAP: 0.8587 -> 0.8593 | Helped |
| Use HSWD instead of SWD for the quantization objective (ablation) | DBDH · COCO, 32 bits | mAP: 0.7996 -> 0.7889 | Hurt |
| Use HSWD quantization rather than the original quantization | Supervised hashing methods · CIFAR-10, NUS-WIDE, and COCO | seconds per epoch: 19.4 -> 17.1; 58.3 -> 50.1; 55.6 -> 49.5 | Helped |
| Use HSWD instead of SWD quantization (ablation) | Supervised hashing methods · CIFAR-10, NUS-WIDE, and COCO | seconds per epoch: 24.2 -> 17.1; 71.2 -> 50.1; 68.1 -> 49.5 | Helped |
Closest papers by meaning
- One Loss for All: Deep Hashing with a Single Cosine Similarity based Learning Objective (NeurIPS 2021 · similarity 0.91)
- Deep Quantization Network for Efficient Image Retrieval (AAAI 2016 · similarity 0.91)
- Deep Hashing Network for Efficient Similarity Retrieval (AAAI 2016 · similarity 0.90)
- Deep Supervised Hashing for Fast Image Retrieval (CVPR 2016 · similarity 0.90)
- Deep Supervised Discrete Hashing (NeurIPS 2017 · similarity 0.90)
- Pairwise-Label-Based Deep Incremental Hashing with Simultaneous Code Expansion (AAAI 2024 · similarity 0.89)
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.