Research map / Label-efficient and robust learning / Domain adaptation and generalization
Burn after Reading: Online Adaptation for Cross-Domain Streaming Data
Luyu Yang, Mingfei Gao, Zeyuan Chen, Ran Xu, Abhinav Shrivastava, Chetan Ramaiah
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 cross-domain bootstrapping with two co-supervised learners (ablation) | single-learner baseline · VisDA-C (online) | online average accuracy: 78.5 -> 79.4 | Improved |
| add cross-domain bootstrapping with two co-supervised learners (ablation) | single-learner baseline · COVID-DA | online average accuracy: 94.4 -> 96.5 | Improved |
| add cross-domain bootstrapping with two co-supervised learners (ablation) | single-learner baseline · WILDS-Camelyon17 | online average accuracy: 86.2 -> 89.2 | Improved |
| add cross-domain bootstrapping with two co-supervised learners (ablation) | single-learner baseline · Fashion-MNIST to DeepFashion | online average accuracy: 42.3 -> 49.1 | Improved |
| add cross-domain bootstrapping with two co-supervised learners (ablation) | single-learner baseline · WILDS-Camelyon17 | one-pass accuracy: 89.4 -> 91.9 | Improved |
| add cross-domain bootstrapping with two co-supervised learners (ablation) | single-learner baseline · Fashion-MNIST to DeepFashion | one-pass accuracy: 39.9 -> 46.3 | Improved |
| add cross-domain bootstrapping with two co-supervised learners (ablation) | single-learner baseline · VisDA-C and COVID-DA | one-pass accuracy: 84.0 -> 84.0 | No clear effect |
| add entropy minimization term (ablation) | single-learner baseline · VisDA-C | online average accuracy: 63.7 -> 78.5 | Improved |
| add class-balancing diversity term (ablation) | single-learner baseline · VisDA-C | one-pass accuracy: 73.0 -> 84.0 | Improved |
| use FixMatch for both learners instead of MixMatch for one (ablation) | CroDoBo · VisDA-C | online average accuracy: 74.3 -> 79.4 | Improved |
| use FixMatch-style l_t instead of MixMatch (ablation) | single-learner baseline · VisDA-C | online average accuracy: 76.3 -> 78.5 | Improved |
| apply RandAugment on entropy and diversity terms (ablation) | single-learner baseline · VisDA-C | online average accuracy: 78.5 -> 77.6 | Hurt |
Closest papers by meaning
- Revisiting Source-Free Domain Adaptation: Insights into Representativeness, Generalization, and Variety (CVPR 2025 · similarity 0.86)
- PrivAuditor: Benchmarking Data Protection Vulnerabilities in LLM Adaptation Techniques (NeurIPS 2024 · similarity 0.84)
- Differentially Private Domain Adaptation with Theoretical Guarantees (ICML 2024 · similarity 0.84)
- Open LLMs are Necessary for Current Private Adaptations and Outperform their Closed Alternatives (NeurIPS 2024 · similarity 0.84)
- A Certified Unlearning Approach without Access to Source Data (ICML 2025 · similarity 0.84)
- The Limits of Differential Privacy in Online Learning (NeurIPS 2024 · similarity 0.83)
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.
