Research map / Graph learning / GNN architectures
DeltaGNN: Graph Neural Network With Information Flow Control
Kevin Mancini, Islem Rekik
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 linear information flow control to DeltaGNN-control (ablation) | DeltaGNN · Cora | accuracy: 84.56 -> 87.29 | Improved |
| add linear information flow control to DeltaGNN-control (ablation) | DeltaGNN · Wisconsin | accuracy: 74.12 -> 80.00 | Improved |
| add linear information flow control to DeltaGNN-control (ablation) | DeltaGNN · Texas | accuracy: 67.57 -> 72.97 | Improved |
| add linear information flow control to DeltaGNN-control (ablation) | DeltaGNN · PubMed | accuracy: 89.64 -> 89.60 | No clear effect |
| add linear information flow control to DeltaGNN-control (ablation) | DeltaGNN · Organ-C | accuracy: 81.44 -> 79.23 | Hurt |
| add linear information flow control to DeltaGNN-control (ablation) | DeltaGNN · Organ-S | accuracy: 62.62 -> 62.09 | Hurt |
| use constant instead of linear control schedule (ablation) | DeltaGNN · Texas | accuracy: 72.97 -> 74.05 | Improved |
| add rewiring with information flow score to GCN (ablation) | GCN · Cornell | accuracy: 37.84 -> 50.54 | Improved |
| add rewiring with information flow score to GCN (ablation) | GCN · Cora | accuracy: 85.12 -> 85.36 | No clear effect |
| add rewiring with information flow score to GAT (ablation) | GAT · Texas | accuracy: 64.32 -> 57.84 | Hurt |
| use IFS instead of other connectivity measures for edge filtering (ablation) | edge filtering · Cora | homophilic rate change: +8.65% | Improved |
| use IFC in DeltaGNN instead of worst-performing model | DeltaGNN · scalability benchmark | epoch time: -30.61% | Improved |
Closest papers by meaning
- From Node Interaction To Hop Interaction: New Effective and Scalable Graph Learning Paradigm (CVPR 2023 · similarity 0.90)
- Bag of Tricks for Training Deeper Graph Neural Networks: A Comprehensive Benchmark Study (TPAMI 2022 · similarity 0.89)
- Long Range Graph Benchmark (NeurIPS 2022 · similarity 0.89)
- Two Heads Are Better Than One: Boosting Graph Sparse Training via Semantic and Topological Awareness (ICML 2024 · similarity 0.88)
- Improving Graph Neural Networks by Learning Continuous Edge Directions (ICLR 2025 · similarity 0.88)
- TopoTune: A Framework for Generalized Combinatorial Complex Neural Networks (ICML 2025 · similarity 0.88)
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.
