View Code? Open in Web Editor
NEW
Thermal Zone modelling using GNN
Python 97.45%
Emacs Lisp 0.17%
TeX 2.38%
thermognn's Introduction
Building Modelling using GNN
- 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
gnn model/training pipeline
data_collect.py
: collects system-excited (randomly set actuation value) -> the data is not in graph format
data_scale.py
: data dimension goes over the scikitlearn.StandardScaler capability
data_process.py
: loop over the collected data, embed the vector data in graph format
gcn2.py
: include PyG dataloader + model + training code