Research map / Label-efficient and robust learning / Domain adaptation and generalization
Maximum Classifier Discrepancy for Unsupervised Domain Adaptation
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 |
|---|---|---|---|
| Align target features by minimizing task-classifier disagreement after maximizing that disagreement in the classifiers | MCD · SVHN→MNIST | accuracy | Helped |
| Increase generator updates from n=2 to n=4 (ablation) | MCD · Digits and traffic-sign transfers | accuracy: 94.2$\pm$2.6 -> 96.2$\pm$0.4; 93.5$\pm$0.4 -> 94.4$\pm$0.3; 92.1$\pm$0.8 -> 94.2$\pm$0.7; 93.1$\pm$1.9 -> 96.5$\pm$0.3; | Helped |
| Increase generator updates from n=2 to n=4 (ablation) | MCD · VisDA | accuracy: 69.7 -> 71.9 | Helped |
| Use gradient reversal layer instead of alternating training steps (ablation) | MCD · GTA5→Cityscapes | mIoU: 38.1 -> 39.9 | Helped |
| Use gradient reversal layer instead of alternating training steps (ablation) | MCD · Synthia→Cityscapes | mIoU: 37.3 -> 34.8 | Hurt |
| Use L2 distance instead of L1 discrepancy (ablation) | MCD · Unspecified experiments | performance | Hurt |
| Omit the source classification loss while training classifiers to maximize discrepancy (ablation) | MCD · Unspecified experiments | performance | Hurt |
| Omit the generator discrepancy-minimization step (ablation) | MCD · Toy inter-twinning moons | target classification | Hurt |
| Increase generator updates from k=2 to k=4 (ablation) | MCD · GTA5→Cityscapes | mIoU: 39.7 -> 38.1 | Mixed |
Closest papers by meaning
- Unsupervised Domain Adaptation With Similarity Learning (CVPR 2018 · similarity 0.91)
- Bridging Theory and Algorithm for Domain Adaptation (ICML 2019 · similarity 0.91)
- Reusing the Task-Specific Classifier as a Discriminator: Discriminator-Free Adversarial Domain Adaptation (CVPR 2022 · similarity 0.90)
- Cross-Domain Gradient Discrepancy Minimization for Unsupervised Domain Adaptation (CVPR 2021 · similarity 0.90)
- Domain-Symmetric Networks for Adversarial Domain Adaptation (CVPR 2019 · similarity 0.90)
- Unsupervised Domain Adaptation Using Feature-Whitening and Consensus Loss (CVPR 2019 · 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.