Research map / Graph learning / Graph self-supervised learning
SAOT: Self-Supervised Continual Graph Learning with Structure-Aware Optimal Transport
Yuting Zhang, Yanbei Liu, Zhitao Xiao, Lei Geng, Yanwei Pang, Xiao Wang
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-task transport-plan distillation (ablation) | SAOT · CoraFull-CL (Class-IL) | AP (%): 72.3 -> 76.3 | Improved |
| add optimal transport alignment instead of Graph Barlow Twins objective (ablation) | SAOT · CoraFull-CL (Class-IL) | AP (%): 71.6 -> 76.3 | Improved |
| distill at transport-plan level instead of cosine embedding constraint (ablation) | SAOT · CoraFull-CL (Class-IL) | AP (%): 68.5 -> 76.3 | Improved |
| distill at transport-plan level instead of MSE embedding constraint (ablation) | SAOT · CoraFull-CL (Class-IL) | AP (%): 69.1 -> 76.3 | Improved |
| add OT alignment and plan distillation under Task-IL (ablation) | SAOT · CoraFull-CL (Task-IL) | AP (%): 95.2 -> 96.7 | Improved |
| increase alpha (intra-task structural calibration) (ablation) | SAOT · CoraFull-CL (Class-IL) | AP (%) | Improved |
| increase alpha (intra-task structural calibration) on large graph (ablation) | SAOT · Reddit-CL (Class-IL) | AP (%) | No clear effect |
| use stronger cross-task regularization beta beyond 0.5-0.6 (ablation) | SAOT · CoraFull-CL and Reddit-CL (Class-IL) | AP / AF | Hurt |
| apply SAOT versus TRACE on large-scale graph | SAOT · Products-CL (Class-IL) | AP (%): 25.6 -> 40.7 | Improved |
| apply SAOT versus TRACE on Arxiv-CL in forgetting | SAOT · Arxiv-CL (Class-IL) | AF (%): -6.4 -> -17.5 | Hurt |
Closest papers by meaning
- Efficient Graph Continual Learning via Lightweight Graph Neural Tangent Kernels-based Dataset Distillation (ICML 2025 · similarity 0.88)
- A Topology-aware Graph Coarsening Framework for Continual Graph Learning (NeurIPS 2024 · similarity 0.87)
- A Selective Learning Method for Temporal Graph Continual Learning (ICML 2025 · similarity 0.87)
- Dynamic Sub-graph Distillation for Robust Semi-supervised Continual Learning (AAAI 2024 · similarity 0.87)
- SSTAG: Structure-Aware Self-Supervised Learning Method for Text-Attributed Graphs (NeurIPS 2025 · similarity 0.86)
- HGOT: Self-supervised Heterogeneous Graph Neural Network with Optimal Transport (ICML 2025 · 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.
