Deep Gaussian Process implementation

Requiements

python 3.6

Requiements:

• gpytorch==1.6.0
• keras==2.6.0
• Keras-Preprocessing==1.1.2
• matplotlib==2.2.5
• numpy==1.17.3
• optuna==3.0.6
• pandas==1.1.5
• plotly==5.15.0
• pyparsing==2.4.7
• scikit-learn==0.24.2
• scipy==1.3.3
• seaborn==0.11.2
• tensorboard==2.10.1
• torch==1.10.2

Usage: HPO

The Hyperparameter optimization (HPO) trial can be started by runing the “experiment.py” file. The change of configuration for this file is

• If run locally:

  hpo = input("Enter HPO Study type: ")    # GP, DGP or DSPP
  pruner_type = input("Enter Pruner Type: ")    # Optuna Pruner Type, either "None" or "HB" for Hyperband
  time_tolerance = int(input("Enter time tolerance: "))   # Maximum number of seconds before cancel a HPO trial

  Enter HPO Study type: DGP
  Enter Pruner Type: None
  Enter time tolerance: 6000
  

• If run on slurm: please change the configurations in the “experiment.py” file.

Usage: file_read.py

The training data and test data extraction process can be done by the “file_read.py” file,

Usage: Model Training

The model training process can be started by runing the “train_GP.py”, “train_DGP.py” and “train_DSPP.py” file.

The trained model state file (*.pth file) and all the model parameters in *.txt file will be saved after training process is finished.

Usage: Result Plot

The model result visulization can be shown by runing the “plot_GP_result_test.py”, “plot_DGP_result_test.py” and “plot_DSPP_result_test.py” file.