Grid Search for GNNs #1070
Description
Enhancement description
In order to find the best GNN architecture suitable for our task, we conduct a grid serach for foru different model types.
Tha parameters that are varied are the model depth (number of layers) and the model width (number of units per layer).
This code is important to ensure replicability of the results of the master thesis.
The training data is generated by the MEmilio graph-ODE model and then split into input and label for the model as implemented in #1057.
- grid serach for four model types (GCNConv, ARMAConv, APPNPConv, GATConv)
Checklist
- Attached labels, especially loc:: or model:: labels.
- Linked to project