Research map / Label-efficient and robust learning / Domain adaptation and generalization
Active Domain Adaptation via Clustering Uncertainty-Weighted Embeddings
Viraj Prabhu, Arjun Chandrasekaran, Kate Saenko, Judy Hoffman
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 |
|---|---|---|---|
| use CLUE instead of BADGE for label acquisition (finetuning) | active DA with finetuning · DomainNet 4-shift average | accuracy: 0.7% | Improved |
| use CLUE instead of BADGE for label acquisition (MME) | active DA with MME · DomainNet 4-shift average | accuracy: 0.8% | Improved |
| use CLUE instead of margin and coreset (finetuning) | active DA with finetuning · DomainNet 4-shift average | accuracy: 1.4% and 3% | Improved |
| use CLUE instead of margin and coreset (MME) | active DA with MME · DomainNet 4-shift average | accuracy: 1.3% and 2.3% | Improved |
| use CLUE instead of margin, coreset, BADGE (finetuning) | active DA with finetuning · SVHN to MNIST | accuracy: 1.9%, 12.3%, and 5.2% | Improved |
| use CLUE+DANN instead of AADA | DANN based active DA · DomainNet | accuracy: 0.4%-2% | Improved |
| use CLUE+DANN instead of AADA on hardest shift | DANN based active DA · DomainNet Clipart to Quickdraw | accuracy: 3.4% | Improved |
| use CLUE+DANN instead of AADA on DIGITS and Office | DANN based active DA · SVHN to MNIST and DSLR to Amazon | accuracy: 2.4% and 7.9% | Improved |
| add MME adaptation on top of finetuning | active DA · DomainNet | accuracy: 2.4-2.7% | Improved |
| use uncertainty-only or diversity-only sampling instead of random | entropy, margin, coreset · hard shifts S to P and C to Q | accuracy | Hurt |
| weight clustering by entropy instead of uniform weights (ablation) | CLUE · DomainNet Clipart to Sketch | accuracy | Improved |
| weight clustering by entropy instead of margin score (ablation) | CLUE · DomainNet Clipart to Sketch | accuracy | No clear effect |
Closest papers by meaning
- Category-Aware Active Domain Adaptation (ICML 2024 · similarity 0.89)
- Active Learning for Domain Adaptation: An Energy-Based Approach (AAAI 2022 · similarity 0.89)
- Composite Active Learning: Towards Multi-Domain Active Learning with Theoretical Guarantees (AAAI 2024 · similarity 0.88)
- Active Domain Adaptation with False Negative Prediction for Object Detection (CVPR 2024 · similarity 0.87)
- Revisiting the Domain Shift and Sample Uncertainty in Multi-source Active Domain Transfer (CVPR 2024 · similarity 0.87)
- Enhancing Semi-supervised Domain Adaptation via Effective Target Labeling (AAAI 2024 · 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.
