Research map / Architectures and efficiency / Quantization
Affinity Preserving Quantization for Hashing: A Vector Quantization Approach to Learning Compact Binary Codes
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 |
|---|---|---|---|
| add affinity preservation term to vector quantization (APQ* to APQ) (ablation) | APQ · NUS-WIDE ITQ 64 bits | mAP: 0.4259 -> 0.4825 | Helped |
| add affinity preservation term to vector quantization (ablation) | APQ · NUS-WIDE ITQ 128 bits | mAP: 0.5189 -> 0.6013 | Helped |
| add affinity preservation term to vector quantization (ablation) | APQ · GIST1M ITQ 128 bits | mAP: 0.5826 -> 0.6571 | Helped |
| add affinity preservation term to vector quantization (ablation) | APQ · NUS-WIDE SIKH 64 bits | mAP: 0.1600 -> 0.1856 | Helped |
| replace scalar quantization (AQ) with baseline vector quantization (APQ*) | hashing quantization · NUS-WIDE ITQ 64 bits | mAP: 0.3476 -> 0.4259 | Helped |
| replace AQ with baseline vector quantization (APQ*) with SIKH projection | hashing quantization · NUS-WIDE SIKH 64 bits | mAP: 0.1625 -> 0.1600 | Hurt |
| replace single-bit quantization with APQ | ITQ hashing · NUS-WIDE 64 bits | mAP: 0.2415 -> 0.4825 | Helped |
| replace OPQ with PCA+APQ | binary coding · ImageNet1M | mAP gain: +1.0%, +1.5%, +4.0%, +8.2% at 16/32/64/128 bits | Helped |
Closest papers by meaning
- One Loss for Quantization: Deep Hashing With Discrete Wasserstein Distributional Matching (CVPR 2022 · similarity 0.89)
- Optimizing affinity-based binary hashing using auxiliary coordinates (NeurIPS 2016 · similarity 0.88)
- Unleashing the Full Potential of Product Quantization for Large-Scale Image Retrieval (NeurIPS 2023 · similarity 0.87)
- A Lower Bound of Hash Codes' Performance (NeurIPS 2022 · similarity 0.86)
- Pairwise-Label-Based Deep Incremental Hashing with Simultaneous Code Expansion (AAAI 2024 · similarity 0.85)
- Qinco2: Vector Compression and Search with Improved Implicit Neural Codebooks (ICLR 2025 · similarity 0.84)
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.