Research map / Label-efficient and robust learning / Domain adaptation and generalization
T-SVDNet: Exploring High-Order Prototypical Correlations for Multi-Source Domain Adaptation
Ruihuang Li, Xu Jia, Jianzhong He, Shuaijun Chen, Qinghua Hu
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 entropy minimization on target domain to Source-Only (ablation) | T-SVDNet · Digits-Five (avg) | accuracy: 86.83 -> 88.19 | Improved |
| add tensor-low-rank constraint (+T) (ablation) | T-SVDNet (+E) · Digits-Five (avg) | accuracy: 88.19 -> 92.68 | Improved |
| add tensor-low-rank constraint (+T) (ablation) | T-SVDNet (+E) · Digits-Five ->mm | accuracy: 73.85 -> 88.76 (14.91%) | Improved |
| add uncertainty-aware weighting (+U) (ablation) | T-SVDNet (+E+T) · Digits-Five (avg) | accuracy: 92.68 -> 93.94 (1.26%) | Improved |
| add uncertainty-aware weighting (+U) (ablation) | T-SVDNet (+E+T) · Digits-Five ->mm | accuracy: 88.76 -> 91.22 | Improved |
| add entropy minimization on target domain (ablation) | T-SVDNet · Digits-Five ->sv | accuracy: 77.76 -> 77.86 | No clear effect |
| train on all data instead of 25000-sample subset | T-SVDNet · Digits-Five (avg) | accuracy: 93.37 -> 93.94 | Improved |
| add tensor-low-rank constraint (compare TNN curves) (ablation) | T-SVDNet · Digits-Five ->mm | tensor nuclear norm: TNN stabilizes at around 46 | Improved |
| add uncertainty-aware weighting to counter negative transfer | T-SVDNet · DomainNet ->qdr | accuracy: 16.5 vs 6.3 (M3SDA) | Improved |
| multi-source T-SVDNet versus best prior MDA method MDDA | T-SVDNet · DomainNet (avg) | accuracy: 43.2 -> 47.0 | Improved |
| T-SVDNet versus MDDA | T-SVDNet · PACS (avg) | accuracy: 86.11 -> 91.25 | Improved |
| T-SVDNet versus M3SDA on ->inf | T-SVDNet · DomainNet ->inf | accuracy: 26.0 -> 25.0 | Hurt |
Later papers that compared with this method
3 later accepted papers report T-SVDNet as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- Subspace Identification for Multi-Source Domain Adaptation (NeurIPS 2023 · better on 1 of 1 shared result)
- Partial disentanglement for domain adaptation (ICML 2022 · better on 1 of 1 shared result)
- Contrastive Vicinal Space for Unsupervised Domain Adaptation (ECCV 2022 · better on 1 of 1 shared result)
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
- STEM: An Approach to Multi-Source Domain Adaptation With Guarantees (ICCV 2021 · similarity 0.89)
- Minimum Class Confusion for Versatile Domain Adaptation (ECCV 2020 · similarity 0.89)
- Subspace Identification for Multi-Source Domain Adaptation (NeurIPS 2023 · similarity 0.88)
- Mind the Class Weight Bias: Weighted Maximum Mean Discrepancy for Unsupervised Domain Adaptation (CVPR 2017 · similarity 0.88)
- Revisiting the Domain Shift and Sample Uncertainty in Multi-source Active Domain Transfer (CVPR 2024 · similarity 0.88)
- Bayesian Domain Adaptation with Gaussian Mixture Domain-Indexing (NeurIPS 2024 · similarity 0.87)
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.
