Research map / Continual learning / Online and replay-based
Self-Supervised Continual Graph Learning in Adaptive Riemannian Spaces
Li Sun, Junda Ye, Hao Peng, Feiyang Wang, Philip S. Yu
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 adaptive curvature rather than restricting the model to hyperspherical space (ablation) | RieGrace · Citeseer and Actor | PM (%): 70.2(1.5) -> 83.6(2.4); 53.4(3.1) -> 61.9(1.2) | Improved |
| Use adaptive curvature rather than restricting the model to Euclidean space (ablation) | RieGrace · Citeseer and Actor | PM (%): 69.8(0.9) -> 83.6(2.4); 52.9(2.7) -> 61.9(1.2) | Improved |
| Use adaptive curvature rather than restricting the model to hyperbolic space (ablation) | RieGrace · Citeseer and Actor | PM (%): 80.9(0.2) -> 83.6(2.4); 56.6(2.4) -> 61.9(1.2) | Improved |
| Use the Lorentz layer instead of tangent-space logarithmic and exponential maps (ablation) | RieGrace · Citeseer and Actor | PM (%): 81.2(1.8) -> 83.6(2.4); 58.5(0.6) -> 61.9(1.2) | Improved |
| Use adaptive curvature and label-free Lorentz distillation | continual node classification · Cora, Citeseer, Actor, ogbn-arXiv, Reddit | PM (%): 95.2(0.8) vs 93.9(0.3); 83.6(2.4) vs 81.5(1.2); 61.9(1.2) vs 60.2(3.3); 83.9(0.3) vs 82.5(0.2); 97.9(1.8) vs 96.7(1.2) | Improved |
| Estimate curvature with CurvNet rather than ComC | ogbn-arXiv graph embeddings · Task Graphs 1–3 | embedding distortion: 0.507 (0.012) -> 0.435 (0.027); 0.653 (0.007) -> 0.490(0.010); 0.524(0.033) -> 0.367 (0.082) | Improved |
Closest papers by meaning
- Adaptive Riemannian Graph Neural Networks (AAAI 2026 · similarity 0.90)
- A Self-Supervised Mixed-Curvature Graph Neural Network (AAAI 2022 · similarity 0.90)
- Motif-Aware Riemannian Graph Neural Network with Generative-Contrastive Learning (AAAI 2024 · similarity 0.89)
- Toward a Unified Geometry Understanding : Riemannian Diffusion Framework for Graph Generation and Prediction (NeurIPS 2025 · similarity 0.88)
- Pseudo-Riemannian Graph Convolutional Networks (NeurIPS 2022 · similarity 0.88)
- Spectro-Riemannian Graph Neural Networks (ICLR 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.
