Research map / Graph learning / Graph self-supervised learning
Data-Centric Learning from Unlabeled Graphs with Diffusion Model
Gang Liu, Eric Inae, Tong Zhao, Jiaxin Xu, Tengfei Luo, Meng Jiang
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 |
|---|---|---|---|
| Generate task-specific labeled augmentations with DCT instead of the strongest baseline | graph property prediction · ogbg-BACE | AUC (%): 82.4(2.4) -> 85.6(0.6) | Improved |
| Generate task-specific labeled augmentations with DCT instead of the strongest baseline | graph property prediction · ogbg-FreeSolv | MAE: 1.547(0.082) -> 1.339(0.075) | Improved |
| Generate task-specific labeled augmentations with DCT instead of the strongest baseline | graph property prediction · GlassTemp | MAE: 26.6(0.3) -> 23.7(0.2) | Improved |
| Add objective I1 to the augmentation process (ablation) | DCT on ogbg-BACE with QM9 unlabeled graphs · ogbg-BACE | AUC (%): 84.4(2.6) -> 85.2(1.3) | Improved |
| Add objective I2 to the augmentation process (ablation) | DCT on ogbg-BACE with QM9 unlabeled graphs · ogbg-BACE | AUC (%): 84.4(2.6) -> 84.7(1.8) | Improved |
| Add both objectives I1 and I2 (ablation) | DCT on ogbg-BACE with QM9 unlabeled graphs · ogbg-BACE | AUC (%): 84.4(2.6) -> 85.6(0.6) | Improved |
| Add both objectives I1 and I2 (ablation) | DCT on ogbg-SIDER with QM9 unlabeled graphs · ogbg-SIDER | AUC (%): 63.7(0.3) -> 63.9(0.3) | Improved |
| Add both objectives I1 and I2 (ablation) | DCT on ogbg-FreeSolv with QM9 unlabeled graphs · ogbg-FreeSolv | MAE: 1.473(0.192) -> 1.339(0.075) | Improved |
| Add both objectives I1 and I2 (ablation) | DCT on O2Perm with QM9 unlabeled graphs · O2Perm | MAE: 177.4(27.3) -> 165.6(24.3) | Improved |
| Replace QM9 with ZINC as the unlabeled source while using both objectives (ablation) | DCT · ogbg-BACE | AUC (%): 85.6(0.6) -> 84.9(0.4) | Hurt |
| Replace QM9 with ZINC as the unlabeled source while using both objectives (ablation) | DCT · ogbg-SIDER | AUC (%): 63.9(0.3) -> 63.7(0.7) | Hurt |
| Replace QM9 with ZINC as the unlabeled source while using both objectives (ablation) | DCT · ogbg-FreeSolv | MAE: 1.339(0.075) -> 1.408(0.092) | Hurt |
Closest papers by meaning
- Cross-Domain Graph Data Scaling: A Showcase with Diffusion Models (NeurIPS 2025 · similarity 0.87)
- Unifying Generation and Prediction on Graphs with Latent Graph Diffusion (NeurIPS 2024 · similarity 0.87)
- SDMG: Smoothing Your Diffusion Models for Powerful Graph Representation Learning (ICML 2025 · similarity 0.87)
- Simple and Critical Iterative Denoising: A Recasting of Discrete Diffusion in Graph Generation (ICML 2025 · similarity 0.85)
- Leveraging Diffusion Model as Pseudo-Anomalous Graph Generator for Graph-Level Anomaly Detection (ICML 2025 · similarity 0.85)
- Generative Modelling of Structurally Constrained Graphs (NeurIPS 2024 · similarity 0.85)
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.
