Research map / Graph learning / Graph self-supervised learning
SPA: A Graph Spectral Alignment Perspective for Domain Adaptation
Zhiqing Xiao, Haobo Wang, Ying Jin, Lei Feng, Gang Chen, Fei Huang, Junbo Zhao
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 graph spectral penalty and pseudo-label loss together (ablation) | CDAN · OfficeHome | avg accuracy: 68.5 -> 75.1 | Improved |
| add graph spectral penalty L_gsa (to L_nap-only model) (ablation) | SPA · OfficeHome | avg accuracy: 71.6 -> 75.1 | Improved |
| add pseudo-labelling loss L_nap (to L_gsa-only model) (ablation) | SPA · OfficeHome | avg accuracy: 72.9 -> 75.1 | Improved |
| vary EMA beta from 0.1 to 0.5 (ablation) | SPA · OfficeHome | avg accuracy: 74.7 -> 75.2 | No clear effect |
| vary EMA beta from 0.5 to 0.9 (ablation) | SPA · OfficeHome | avg accuracy: 75.2 -> 74.8 | No clear effect |
| use Gaussian instead of cosine similarity (ablation) | SPA L_rwk · OfficeHome | avg accuracy: 74.7 -> 75.3 | Improved |
| use random walk instead of symmetric Laplacian (ablation) | SPA · OfficeHome | avg accuracy: 75.1 -> 75.3 | No clear effect |
| increase kNN k from 3 to 5 (ablation) | SPA L_rwk · OfficeHome | avg accuracy: 74.7 -> 75.3 | Improved |
| increase kNN k from 3 to 5 (ablation) | SPA L_sym · OfficeHome | avg accuracy: 74.8 -> 75.1 | Improved |
| apply SPA on DANN vs MixMatch for VisDA2017 (ablation) | SPA · VisDA2017 | accuracy: 83.7 -> 87.7 | Improved |
| add SPA to DANN baseline (ablation) | DANN · Office31 | avg accuracy: 82.2 -> 91.4 | Improved |
| add SPA to DANN baseline (ablation) | DANN · OfficeHome | avg accuracy: 57.6 -> 75.3 | Improved |
Closest papers by meaning
- Open-Set Graph Domain Adaptation via Separate Domain Alignment (AAAI 2024 · similarity 0.90)
- Homophily Enhanced Graph Domain Adaptation (ICML 2025 · similarity 0.90)
- Disentangled Graph Spectral Domain Adaptation (ICML 2025 · similarity 0.89)
- Revisiting, Benchmarking and Understanding Unsupervised Graph Domain Adaptation (NeurIPS 2024 · similarity 0.89)
- Towards Unsupervised Open-Set Graph Domain Adaptation via Dual Reprogramming (NeurIPS 2025 · similarity 0.89)
- TFGDA: Exploring Topology and Feature Alignment in Semi-supervised Graph Domain Adaptation through Robust Clustering (NeurIPS 2024 · similarity 0.88)
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.
