Research map / Label-efficient and robust learning / Recognition and long-tail
MIHash: Online Hashing With Mutual Information
Fatih Cakir, Kun He, Sarah Adel Bargal, Stan Sclaroff
Read the paperCheck your idea against 137,217 papers
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 mutual information Trigger Update to OKH | OKH · CIFAR-10 online (32 bits) | AUC of mAP: 0.259 -> 0.287 | Improved |
| add Trigger Update to OKH | OKH · LabelMe online (32 bits) | AUC of mAP: 0.198 -> 0.374 | Improved |
| add Trigger Update to AdaptHash | AdaptHash · CIFAR-10 online (32 bits) | AUC of mAP: 0.218 -> 0.250 | Improved |
| add Trigger Update to OSH | OSH · CIFAR-10 online (32 bits) | AUC of mAP: 0.463 -> 0.466 | No clear effect |
| add Trigger Update to OSH | OSH · Places205 online (32 bits) | AUC of mAP: 0.246 -> 0.236 | Hurt |
| add Trigger Update to OKH | OKH · LabelMe online (32 bits) | hash table updates: reduced by a factor of 67 | Improved |
| raise TU threshold theta from 0 to at least 0.2 (ablation) | OKH · CIFAR-10 online (32 bits) | AUC of mAP: 0.287 -> 0.134 | Hurt |
| raise TU threshold theta from 0 to at least 0.1 (ablation) | OKH · Places205 online (32 bits) | AUC of mAP: 0.161 -> 0.014 | Hurt |
| train MIHash for full 100 epochs instead of one epoch (ablation) | MIHash · CIFAR-10 supervised Setting 1 (48 bits) | mAP: 0.609 -> 0.746 | Improved |
| use mutual information objective instead of pairwise or triplet objectives | MIHash vs DTSH · CIFAR-10 supervised Setting 1 (48 bits) | mAP: 0.702 -> 0.746 | Improved |
| use mutual information objective with end-to-end VGG-F | MIHash vs DTSH · CIFAR-10 supervised Setting 2 (48 bits) | mAP: 0.934 -> 0.942 | Improved |
| optimize mutual information with MIHash from random hash mapping | MIHash · CIFAR-10 test, 20K online | mAP: 0.22 -> 0.68 | Improved |
Later papers that compared with this method
4 later accepted papers report MIHash as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.
- Fast Online Hashing with Multi-Label Projection (AAAI 2023 · better on 3 of 3 shared results)
- Hadamard Matrix Guided Online Hashing (IJCV 2020 · better on 5 of 5 shared results)
- Fast Class-wise Updating for Online Hashing (TPAMI 2020 · better on 3 of 4 shared results)
- Hashing with Binary Matrix Pursuit (ECCV 2018 · better on 13 of 18 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Hashing with Mutual Information (TPAMI 2019 · similarity 0.92)
- A Survey on Learning to Hash (TPAMI 2017 · similarity 0.91)
- Towards Optimal Discrete Online Hashing with Balanced Similarity (AAAI 2019 · similarity 0.91)
- An ensemble diversity approach to supervised binary hashing (NeurIPS 2016 · similarity 0.90)
- Optimizing affinity-based binary hashing using auxiliary coordinates (NeurIPS 2016 · similarity 0.90)
- Fast Online Hashing with Multi-Label Projection (AAAI 2023 · similarity 0.90)
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.
