Lots of learning tasks require dealing with graph data which contains rich relation information among elements. Modeling physics systems, learning molecular fingerprints, predicting protein interface, and classifying diseases demand a model to learn from graph inputs. In other domains such as learning from non-structural data like texts and images, reasoning on extracted structures (like the dependency trees of sentences and the scene graphs of images) is an important research topic which also needs graph reasoning models. Graph neural networks (GNNs) are neural models that capture the dependence of graphs via message passing between the nodes of graphs. In recent years, variants of GNNs such as graph convolutional network (GCN), graph attention network (GAT), graph recurrent network (GRN) have demonstrated ground-breaking performances on many deep learning tasks.
| Authors | Zhou, J.; Cui, G.; Hu, S.; Zhang, Z.; Yang, C.; Liu, Z.; Sun, M. |
|---|---|
| Year | 2020 |
| Venue | AI Open 1, 57–81 |
| DOI | 10.1016/j.aiopen.2021.01.001 |
| Source Database | Backtracking |
| Bridge-to-GNN | Category E |
| GNN Architecture | DGCNN (heuristic — verify) |
| Graph Encoding | Not yet extracted from PDF |
| AEC Task | Not yet extracted from PDF |
| Cohort | Early Infrastructure (2020–2021) |
| Dataset Size | 20,000 nodes (heuristic — verify) |
|---|---|
| Implementation Framework | PyTorch Geometric (PyG), DGL, PyTorch, Topologic (heuristic — verify) |
| Key Hyperparameters | 4 layers (heuristic — verify) |
| Primary Metric | Mean Average Precision (mAP) (heuristic — verify) |
| Key Finding | general design pipeline for GNN models and discuss the variants of each component, systematically categorize the applications, and propose four open problems for future research (heuristic — verify) |
| Quality Assessment | 4/6 · rigor: Medium · Code/Data: ✓Benchmark: ✓Baseline: ✓Reproducible: ✓ (heuristic — verify) |
| GNN architectures detected | DGCNN · GraphSAGE / SAGEConv · GAT (Graph Attention) · R-GCN (Relational GCN) · Heterogeneous GNN / HAN · MPNN · Graph Transformer / GraphGPS / SAN · PointNet / PointNet++ · Graph U-Net / DiffPool · ClusterGCN · ChebNet (Chebyshev) · T-GCN / Temporal GCN · Graph VAE / VGAE · Graph Diffusion · Knowledge Graph Embedding (TransE/RotatE/DistMult) · Node2Vec / DeepWalk · GraphSAINT · PinSAGE · GCN |
|---|---|
| Frameworks / libraries | PyTorch Geometric (PyG) · Deep Graph Library (DGL) · PyTorch · MXNet / PaddlePaddle · Topologic / Topologicpy |
| Benchmark datasets referenced | Cora · Citeseer · PubMed · MUTAG · OGB · Open Graph Benchmark · Reddit · PPI · PROTEINS · ENZYMES |
| Code repositories found in text | https://github.com/rusty1s/pytorch_geometric · https://github.com/dmlc/dgl · https://github.com/alibaba/aligraph · https://github.com/DeepGraphLearning/graphvite · https://github.com/PaddlePaddle/PGL |
| Hyperparameters (regex-detected) | layers=4 |
| Dataset stats found | 20,000 nodes |
| Quality (heuristic, 6-flag) | 3/6 · rigor: Medium Code/Data: ✓Benchmark: ✓Baseline: ✓CV/Split: —Ablation: —Reproducible: — |
Part of the GML/GNN in AEC Systematic Review (PRISMA 2020) — 112 papers, 2020–2026