Awesome

Auto-STGCN

An automated system for STGCN model development.<br> Code for paper 'Auto-STGCN: Autonomous Spatial-Temporal Graph Convolutional Network Search Based on Reinforcement Learning and Existing Research Results'.<br>

1. Auto-STGCN Algorithm: Searching for the optimal STGCN model

Related Files

Auto_STGCN.py --- run Auto-STGCN algorithm<br> Model.py --- build STGCN model according to code<br> Env.py --- read dataset, record the state-action-reward information in Auto-STGCN algorithm<br> ExperimentDataLogger.py --- output the log information of Auto-STGCN algorithm<br> /Log --- log files<br> /utils --- auxiliary files<br> /data --- datasets<br> /Config --- default configurations<br>

Inputs Details

• Dataset name, Dataset partition ratio (validation set, test set, training set), Input sequence length, Output sequence length,<br>
• Timemax, Epoch size of each candidate model,<br>
• Initial epsilon, Epsilon decay step, Epsilon decay Ratio, Gamma of Qlearning, Learning rate of Qlearning, Episodes of Qlearning<br>

Outputs Details

• Code and performance scores of the Optimal STGCN searched by Auto-STGCN<br>
• Log info of Auto-STGCN<br>

Commands

• python Auto_STGCN.py --data "PEMS03"<br>
• python Auto_STGCN.py --data "PEMS03" --gamma 0.1<br>

2. Auto-STGCN Algorithm: Training the optimal STGCN model

Related Files

TestBestGNN.py --- train the optimal STGCN model searched by Auto-STGCN algorithm<br> Model.py --- build STGCN model according to code<br> /Log --- log files<br> /utils --- auxiliary files<br> /data --- datasets<br> /Config --- default configurations<br>

Inputs Details

• Optimal STGCN code, Dataset name, Dataset partition ratio (validation set, test set, training set), Input sequence length, Output sequence length,<br>
• Model training epochs, Model training times,<br>
• Load model weight = None<br>

Outputs Details

• Performance scores (Mean + variance: MAE, MAPE, RMSE, Time) of the Optimal STGCN model<br>
• Log info of the model training<br>

Commands

• python TestBestSTGNN.py --model "./Config/qlearning_2.json" --data "PEMS03"<br>
• python TestBestSTGNN.py --model "./Config/qlearning_2.json" --data "PEMS03" --gamma 0.1<br>

3. Auto-STGCN Algorithm: Loading the optimal STGCN model

Related Files

TestBestGNN.py --- test the performance of optimal STGCN model searched by Auto-STGCN algorithm<br> Model.py --- build STGCN model according to code<br> /Log --- log files<br> /utils --- auxiliary files<br> /data --- datasets<br> /Config --- default configurations<br>

Inputs Details

• Dataset name, test number, Load model weight = Model loading path <br>

Outputs Details

• Performance scores (Mean + variance: MAE, MAPE, RMSE, Time) of the Optimal STGCN model on test set<br>

Commands

• python TestBestGNN.py --data "PEMS03" --load "./Log/PEMS03_experiment2_qlearning2_test/GNN/best_GNN_model.params" --times 1<br>