How georgecao-hg/gnn-node-prediction-example works
This is a runnable node-level GNN regression demo. The principal workflow generates synthetic graphs, splits and batches them, trains a two-layer model, and evaluates it each epoch. The separate simplified convolution file demonstrates neighbor aggregation but is not used by the main script. The graph follows the shown calls and README workflow; it does not assume that the standalone implementation participates in training.
Main entry points: main.py Demo entry point
Components 10
Data and training
- Demo entry point
main.py - Synthetic graphs
main.py - Split and batch
main.py - Training epoch
main.py - Test evaluation
main.py
Graph models
- Node prediction model
main.py - PyG GCNConv
- Simplified convolution
SimplifiedGCNConv.py
Other components
- User
- Console output
Connections 10
- User → Demo entry point: runs script
readme.md - Demo entry point → Synthetic graphs: generates graphs
main.py - Demo entry point → Split and batch: splits and batches
main.py - Demo entry point → Node prediction model: initializes
main.py - Demo entry point → Training epoch: runs each epoch
main.py - Demo entry point → Test evaluation: evaluates each epoch
main.py - Training epoch → Node prediction model: predicts with
main.py - Test evaluation → Node prediction model: predicts with
main.py - Demo entry point → Console output: prints losses
main.py - Node prediction model → PyG GCNConv: applies layers
main.py