Experiment Settings

class supernnova.utils.experiment_settings.ExperimentSettings(cli_args)

Bases: object

Mother class to control experiment parameters

This class is responsible for the following

• Defining paths and model names

• Choosing the device on which to run computations

• Specifying all hyperparameters such as model configuration, datasets, features etc

Parameters:

cli_args (argparse.Namespace) –

get_randomforest_features()

Specify list of features to be used for RandomForest training

setup_dir()

Configure directories where data is read from or dumped to during the course of an experiment

set_pytorch_model_name()

Define the model name for all NN based classifiers

set_randomforest_model_name()

Define the model name for all RandomForest based classifiers

check_data_exists()

Utility to check the database has been built

set_feature_lists()

Utility to define the features used to train NN=based models

set_database_file_names()

Create a unique database name based on the dataset required by the settings

load_normalization()

Create an array holding the data-normalization parameters used to normalize certain features in the NN-based classification pipeline