- the most accurate model is defined in file
gcn2.py
- uses GNN-RNN architecture and was able to achieve on par performance to a typical mlp (RNN) based model
- gcn2 uses a GCN for generating a representation of the graph nodes, and the nodes are passed into a RNN layer
data_collect.py
: collects system-excited (randomly set actuation value) -> the data is not in graph formatdata_scale.py
: data dimension goes over the scikitlearn.StandardScaler capabilitydata_process.py
: loop over the collected data, embed the vector data in graph formatgcn2.py
: include PyG dataloader + model + training code