Research map / Learning theory and algorithms / Prediction and decision losses
Optimizing affinity-based binary hashing using auxiliary coordinates
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 |
|---|---|---|---|
| Apply MAC (auxiliary coordinates, iterated Z and h steps) on top of cut codes instead of two-step learning | KSH loss with linear and kernel SVM hash functions · CIFAR, 16-48 bits | KSH loss value and precision/recall | Helped |
| Apply MAC on top of quad codes instead of two-step learning | KSH loss with linear and kernel SVM hash functions · CIFAR, 16-48 bits | KSH loss value and precision/recall | Helped |
| Use cut (GraphCut) rather than quad (quadratic surrogate) for the Z step in MAC (ablation) | MAC with KSH loss · CIFAR | precision and loss | Helped |
| Apply MAC to eSPLH loss instead of cut only | eSPLH loss, MACcut · CIFAR, linear and kernel SVM | loss and precision/recall | Helped |
| Use MAC with a fixed-schedule penalty plus validation early stopping | MAC algorithm · CIFAR | validation precision | No clear effect |
| MAC on top of cut/quad vs. comparing against state-of-the-art hashing methods (ITQ, BRE, KSH, STH) | MACcut and MACquad · CIFAR and Infinite MNIST, linear hash, 16-64 bits | precision and precision/recall | Helped |
| Apply MAC on SIFT1M unsupervised-like pseudolabel setting, MACquad vs MACcut (ablation) | MAC with KSH loss, kappa+=50, kappa-=1000, K=10000 · SIFT1M | precision and precision/recall | Mixed |
| Use MAC rather than two-step; runtime cost | MACcut / MACquad · CIFAR 10,000 points, 48 bits | runtime per iteration | No clear effect |
Closest papers by meaning
- An ensemble diversity approach to supervised binary hashing (NeurIPS 2016 · similarity 0.91)
- Optimal Projection Guided Transfer Hashing for Image Retrieval (AAAI 2019 · similarity 0.89)
- Deep Supervised Discrete Hashing (NeurIPS 2017 · similarity 0.89)
- On Trivial Solution and High Correlation Problems in Deep Supervised Hashing (AAAI 2018 · similarity 0.88)
- Learning Attribute-Aware Hash Codes for Fine-Grained Image Retrieval via Query Optimization (ICML 2025 · similarity 0.87)
- Generalized Debiased Semi-Supervised Hashing for Large-Scale Image Retrieval (AAAI 2025 · similarity 0.87)
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.