Research map / Graph learning / Graph self-supervised learning
Auto-Encoding Twin-Bottleneck Hashing
Yuming Shen, Jie Qin, Jiaxin Chen, Mengyang Yu, Li Liu, Fan Zhu, Fumin Shen, Ling Shao
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 |
|---|---|---|---|
| add a continuous bottleneck for decoding alongside the binary bottleneck (ablation) | TBH · CIFAR-10, 16/32/64 bits | MAP@1000: 0.435 -> 0.532; 0.437 -> 0.573; 0.433 -> 0.578 | Improved |
| keep binary codes for graph construction and continuous variables for decoding rather than swapping bottleneck roles (ablation) | TBH · CIFAR-10, 16/32/64 bits | MAP@1000: 0.466 -> 0.532; 0.471 -> 0.573; 0.475 -> 0.578 | Improved |
| use WAE adversarial regularizers instead of explicit entropy and L2 regularization (ablation) | TBH · CIFAR-10, 16/32/64 bits | MAP@1000: 0.524 -> 0.532; 0.559 -> 0.573; 0.560 -> 0.578 | Improved |
| add regularization for the binary and continuous bottlenecks (ablation) | TBH · CIFAR-10, 16/32/64 bits | MAP@1000: 0.521 -> 0.532; 0.535 -> 0.573; 0.547 -> 0.578 | Improved |
| use the discrete stochastic neuron for the binary bottleneck instead of removing it and adding quantization loss (ablation) | TBH · CIFAR-10, 16/32/64 bits | MAP@1000: 0.408 -> 0.532; 0.412 -> 0.573; 0.463 -> 0.578 | Improved |
| use an adaptive graph computed from code Hamming distances instead of a fixed feature-distance graph (ablation) | TBH · CIFAR-10, 16/32/64 bits | MAP@1000: 0.442 -> 0.532; 0.464 -> 0.573; 0.459 -> 0.578 | Improved |
| use a GCN rather than weighted averaging for graph-based representation mixing (ablation) | TBH · CIFAR-10, 16/32/64 bits | MAP@1000: 0.477 -> 0.532; 0.503 -> 0.573; 0.519 -> 0.578 | Improved |
Closest papers by meaning
- Structured Learning of Binary Codes with Column Generation for Optimizing Ranking Measures (IJCV 2017 · similarity 0.89)
- One Loss for All: Deep Hashing with a Single Cosine Similarity based Learning Objective (NeurIPS 2021 · similarity 0.89)
- Unsupervised Deep Hashing with Similarity-Adaptive and Discrete Optimization (TPAMI 2018 · similarity 0.88)
- Hashing with Binary Matrix Pursuit (ECCV 2018 · similarity 0.88)
- Pairwise-Label-Based Deep Incremental Hashing with Simultaneous Code Expansion (AAAI 2024 · similarity 0.86)
- Statistical Model-driven Similarity Hashing: Bridging Modalities for Efficient Unsupervised Retrieval (AAAI 2025 · similarity 0.86)
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.
