Research map / Graph learning / Graph self-supervised learning
Confidence-Based Feature Imputation for Graphs with Partially Known Features
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Replace FP's symmetric transition matrix with row-stochastic transition matrix (ablation) | feature propagation (CiteSeer, GCN) · CiteSeer structural missing rm=0.995 | accuracy: +5.04 (59.76 to 64.80) | Helped |
| Add pseudo-confidence channel-wise inter-node diffusion (ablation) | PCFI (CiteSeer, GCN) · CiteSeer structural missing rm=0.995 | accuracy: +0.60 (64.80 to 65.40) | Helped |
| Add node-wise inter-channel propagation (ablation) | PCFI (CiteSeer, GCN) · CiteSeer structural missing rm=0.995 | accuracy: +0.78 (65.40 to 66.18) | Helped |
| Add pseudo-confidence channel-wise inter-node diffusion (ablation) | PCFI (Cora, GAE) · Cora link prediction structural missing rm=0.995 | AUC: +0.20 (83.96 to 84.16) | Helped |
| Add node-wise inter-channel propagation (ablation) | PCFI (Cora, GAE) · Cora link prediction structural missing rm=0.995 | AUC: +2.29 (84.16 to 86.45) | Helped |
| Add node-wise inter-channel propagation and pseudo-confidence diffusion over FP baseline (ablation) | PubMed link prediction (GAE) · PubMed uniform missing rm=0.995 (ablation, Table 7) | AUC: 77.05 to 85.26 across all components | Helped |
| Full PCFI replaces FP (pseudo-confidence imputation with channel-wise diffusion and inter-channel propagation) | node classification with GCN · six datasets, structural missing rm=0.995 | average accuracy: +2.41 (72.92 to 75.33) | Helped |
| Full PCFI replaces FP for node classification | node classification with GCN · six datasets, uniform missing rm=0.995 | average accuracy: +1.27 (76.22 to 77.49) | Helped |
| Full PCFI replaces FP for link prediction (GAE) | link prediction · Cora/CiteSeer/Photo/Computers, structural and uniform missing rm=0.995 | AUC: Cora structural +15.48 (68.26 to 83.74) | Helped |
| Full PCFI vs FP on PubMed link prediction | PubMed link prediction structural missing · PubMed structural missing rm=0.995 | AUC: 78.93 vs 87.14 for FP | Hurt |
Closest papers by meaning
- Propagate and Inject: Revisiting Propagation-Based Feature Imputation for Graphs with Partially Observed Features (ICML 2025 · similarity 0.92)
- Attribute-Missing Graph Clustering Network (AAAI 2024 · similarity 0.85)
- Accurate Link Prediction for Edge-Incomplete Graphs via PU Learning (AAAI 2025 · similarity 0.84)
- Robust Node Classification on Graph Data with Graph and Label Noise (AAAI 2024 · similarity 0.83)
- Missing Data Imputation by Reducing Mutual Information with Rectified Flows (NeurIPS 2025 · similarity 0.83)
- No Metric to Rule Them All: Toward Principled Evaluations of Graph-Learning Datasets (ICML 2025 · similarity 0.83)
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.