Research map / Graph learning / Graph self-supervised learning
MIGDiff: Multi-attributes Imputations for Attribute-missing Graphs via Graph Denoising Diffusion Model
Ye Liu, Yang Chen, Hongmin Cai
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 structural (topological) embeddings (ablation) | MIGDiff · Cora (0.4 observed rate) | Recall@10 (%): 13.82 -> 17.65 | Improved |
| add attribute embeddings (ablation) | MIGDiff · Cora (0.4 observed rate) | Recall@10 (%): 17.37 -> 17.65 | Improved |
| use graph auto-encoder with both embeddings instead of a plain MLP (ablation) | MIGDiff · Cora (0.4 observed rate) | Recall@10 (%): 10.78 -> 17.65 | Improved |
| add VAE-style multi-imputation to VAE | VAE baseline · Cora (0.4 observed rate) | Recall@10 (%): 8.84 -> 12.44 | Improved |
| add VAE-style multi-imputation to SAT | SAT · Cora (0.4 observed rate) | Recall@10 (%): 16.29 -> 12.32 | Hurt |
| add VAE-style multi-imputation to ITR | ITR · Cora (0.4 observed rate) | Recall@10 (%): 16.40 -> 15.46 | Hurt |
| use diffusion generation instead of VAE-based multi-imputation | multi-attribute imputation · Cora diversity, k=5 | average variance: 0.0378 -> 2.9455 | Improved |
| expand graph with three imputed attribute graphs | MIGDiff · node classification, GCN (four datasets) | accuracy | Improved |
| use MIGDiff over best single-attribute baseline with MLP classifier | MIGDiff · Citeseer node classification, MLP | accuracy (%): 68.91 -> 68.73 | No clear effect |
| use MIGDiff at low observed rate (0.2) | MIGDiff · Cora, observed rate 0.2 | Recall/NDCG | No clear effect |
| increase number of imputation samples k (ablation) | MIGDiff · Cora diversity | average variance: 1.8388 -> 2.9455 | Improved |
Closest papers by meaning
- Attribute-Missing Graph Clustering Network (AAAI 2024 · similarity 0.90)
- Propagate and Inject: Revisiting Propagation-Based Feature Imputation for Graphs with Partially Observed Features (ICML 2025 · similarity 0.88)
- Attribute-Missing Multi-view Graph Clustering (CVPR 2025 · similarity 0.86)
- When Edge Independence Fails: Joint Graph Diffusion with Latent Sociability Priors (NeurIPS 2026 · similarity 0.86)
- Learning What Not to Impute: An Uncertainty-Aware Diffusion Framework for Meaningful Missingness (NeurIPS 2026 · similarity 0.86)
- SDMG: Smoothing Your Diffusion Models for Powerful Graph Representation Learning (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.
