training_creator_noiseadd_breast_cancer.py

from datetime import datetime
from model_train import model_train
from model_train_baseline import model_train as model_train_baseline


time_now = datetime.now()
current_time_str = time_now.strftime("%Y%m%d_%H%M")

model_properties = {'architecture': 'modulate_layer_network_compdrop',
                    'number_of_layers': 2,
                    'size_of_layers': [4, 2],
                    'compensation_layer_location': 0,
                    'activation_function': 'relu',
                    'dropout_rate': 0.5,
                    'compensation_architecture': [8, 8],
                    'compensation_activation': 'relu',
                    'compensation_dropout': 0.0,
                    'plus': False,
                    'nan_input': False,
                    'data_input_modulate': True}

training = dict()
training['name'] = 'sklearn_breastcancer_noiseadd_10vars'
training['experiment_name'] = training['name'] + '_' + current_time_str
training['description'] = 'This is a model of for predicting breast cancer from sklearn data'
training['data_location'] = 'sklearn_breastcancer_random_noise_010_20210527_0101'
training['use_gpu'] = True
training['inputs'] = ['mean radius', 'mean texture', 'mean perimeter', 'mean area', 'mean smoothness',
                      'mean compactness', 'mean concavity', 'mean concave points', 'mean symmetry',
                      'mean fractal dimension']
training['outputs'] = ['target']
training['number_of_epochs'] = 50
training['batch_size'] = 64
training['verbose'] = True
training['seed_value'] = 42
training['plotting'] = False

preprocessing = dict()
preprocessing['load_nan'] = True
preprocessing['preop_missing_imputation'] = None
preprocessing['imbalance_compensation'] = 'none'
preprocessing['numerical_inputs'] = ['mean radius', 'mean texture', 'mean perimeter', 'mean area', 'mean smoothness',
                                     'mean compactness', 'mean concavity', 'mean concave points', 'mean symmetry',
                                     'mean fractal dimension']
preprocessing['categorical_inputs'] = []
preprocessing['outputs'] = ['target']
preprocessing['miss_augment'] = None
preprocessing['miss_introduce'] = None
preprocessing['data_subset'] = 1.0

model_parameters = dict()
model_parameters['counter'] = '0'
model_parameters['name'] = training['name']
model_parameters['optimizer'] = {'name': 'SGD',
                                 'learning_rate': 0.1,
                                 'momentum': 0.9,
                                 'weight_decay': 8e-5}
model_parameters['cost_function'] = 'cross_entropy'
model_parameters['model_construction'] = model_properties
model_parameters['loss_weights'] = [1.0, 1.0]
model_parameters['time_series_length'] = []
model_parameters['lr_scheduler'] = None
if model_properties['architecture'] == 'xgboost':
    model_train_baseline(model_parameters, training, preprocessing)
else:
    model_train(model_parameters, training, preprocessing)