Research map / Graph learning / GNN architectures
Adaptive Initial Residual Connections for GNNs with Theoretical Guarantees
Mohammad Shirzadi, Ali Safarpoor Dehkordi, Ahad N. Zehmakan
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 |
|---|---|---|---|
| replace static initial residual (GCNII) with PageRank-based adaptive residual strengths | GNN · Texas, Wisconsin, Cornell, Chameleon, Squirrel (heterophilic) | test accuracy: 59.5 -> 77.0 (Texas); 60.4 -> 79.0 (Wisconsin); 36.2 -> 65.0 (Chameleon); 28.8 -> 49.0 (Squirrel) | Improved |
| adaptive IRC (PageRank) vs GCNII | GNN · Actor | test accuracy: 35.2 -> 33.8 | Hurt |
| use fixed PageRank-based residual strengths instead of learned ones (ablation) | adaptive IRC · Cora, CiteSeer, Pubmed, Texas, Cornell, Chameleon, Squirrel | test accuracy: 80.1 -> 80.7 (Cora); 69.3 -> 70.2 (CiteSeer); 47.7 -> 49.0 (Squirrel) | Improved |
| use fixed PageRank-based residual strengths instead of learned ones (ablation) | adaptive IRC · Wisconsin, Actor | test accuracy: 82.4 -> 79.0 (Wisconsin); 34.2 -> 33.8 (Actor) | Hurt |
| adaptive IRC vs DirGNN | GNN · Texas | test accuracy: 84.6 -> 77.0 | Hurt |
| increase depth beyond four layers with adaptive IRC (ablation) | adaptive IRC · node classification, all datasets | test accuracy | Improved |
| adaptive IRC vs GCN/GAT/GraphSAGE on oversmoothing | adaptive IRC · synthetic SBM, 200 nodes | Dirichlet energy | Improved |
Closest papers by meaning
- Residual Connections and Normalization Can Provably Prevent Oversmoothing in GNNs (ICLR 2025 · similarity 0.90)
- Deep Graph Neural Networks via Posteriori-Sampling-based Node-Adaptative Residual Module (NeurIPS 2024 · similarity 0.89)
- On Vanishing Gradients, Over-Smoothing, and Over-Squashing in GNNs: Bridging Recurrent and Graph Learning (NeurIPS 2025 · similarity 0.88)
- Locality-Aware Graph Rewiring in GNNs (ICLR 2024 · similarity 0.87)
- Accurate and Scalable Graph Neural Networks via Message Invariance (ICLR 2025 · similarity 0.87)
- Towards Dynamic Message Passing on Graphs (NeurIPS 2024 · similarity 0.87)
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.
