## DEFINING INPUT LOCATION
if want_full_path is False:
input_dir_path = os.path.join(input_trajectory_location,
solute, directory_name)
else:
input_dir_path = os.path.join(input_trajectory_location,
directory_name)
print("Input dir path: %s" % (input_dir_path))
## CHECKING IF DIRECTORY PATH EXISTS
if os.path.exists(input_dir_path) is True:
## DEFINING NAME FOR PICKLE
pickle_name = convert_to_single_name(
solute=solute,
solvent=solvents,
mass_fraction=mass_frac,
temp=temperature)
## DEFINING PICKLE PATH
pickle_file_path = os.path.join(output_database_location,
pickle_name + picklesuffix)
### LOADING TRAJECTORY
traj_data = import_tools.import_traj(
directory=input_dir_path, # Directory to analysis
structure_file=gro_file, # structure file
xtc_file=xtc_file, # trajectories
)
## RUNNING GRID INTERPOLATION
def load_xy_data(self):
'''
This function loads all desired xy data.
'''
## READING CSV FILE
csv_file = pd.read_csv(self.class_file_path)
## STORE X DATA AND ITS LABEL (POSITIVE OR NEGATIVE)
self.x_data = []
self.y_label = []
## STORING INSTANCE INFORMATION
self.instance_names = []
## LOOPING THROUGH SOLUTES
for solute in self.solute_list:
## LOOPING THROUGH COSOLVENT
for cosolvent in self.solvent_list:
## LOOPING THROUGH MASS FRACTION OF WATER
for mass_frac in self.mass_frac_data:
## SPECIFYING SPECIFIC TRAINING INSTANCE
training_instance = {
'solute': solute,
'cosolvent': cosolvent,
'mass_frac': mass_frac, # mass fraction of water
'temp': SOLUTE_TO_TEMP_DICT[solute], # Temperature
}
## CONVERTING TRAINING INSTANCE NAME TO NOMENCLATURE
training_instance_name = convert_to_single_name(
solute=training_instance['solute'],
solvent=training_instance['cosolvent'],
mass_fraction=training_instance['mass_frac'],
temp=training_instance['temp'])
##############################
### EXTRACTING CLASS VALUE ###
##############################
## FINDING INSTANCE VALUE
class_instance_value = locate_test_instance_value(
csv_file=csv_file,
solute=training_instance['solute'],
cosolvent=training_instance['cosolvent'],
mass_frac_water=training_instance['mass_frac'],
)
## PRINTING
if (str(class_instance_value) != 'nan'):
self.y_label.append(class_instance_value)
else:
print(
"Training instance: %s, Class value: %s -- skipping!"
% (training_instance_name, class_instance_value))
################################
### EXTRACTING TRAINING SETS ###
################################
if (str(class_instance_value) != 'nan'):
## DEFINING FULL TRAINING PATH
full_train_pickle_path = os.path.join(
self.database_path, training_instance_name)
## PRINTING
print(full_train_pickle_path)
## EXTRACTION PROTOCOL A PARTICULAR TRAINING EXAMPLE
training_data_for_instance = load_pickle(
full_train_pickle_path)
## CHANGING TRAINING DATA INSTANCE REPRESENTATION
training_data_representation, str_output = combine_training_data(
training_data_for_instance=
training_data_for_instance,
representation_type=self.representation_type,
representation_inputs=self.representation_inputs)
## PRINTING
if self.verbose == True:
if str_output is None:
print("Instance: %s, Class value: %s" %
(training_instance_name,
class_instance_value)
) # Should output negative
else:
print("Instance: %s, Class value: %s, %s" %
(training_instance_name,
class_instance_value,
str_output)) # Should output negative
## STORING
self.x_data.append(training_data_representation)
self.instance_names.append(training_instance_name)
## FINDING TOTAL INSTANCES
self.total_instances = len(self.instance_names)
return
(class_instance_value)) # Should output negative
# if (class_instance_value == 'positive'):
# y_label.append(1)
# elif (class_instance_value == 'negative'):
# y_label.append(0)
if (str(class_instance_value) != 'nan'):
y_label.append(class_instance_value)
y_label.append(class_instance_value)
y_label.append(class_instance_value)
y_label.append(class_instance_value)
y_label.append(class_instance_value)
## CONVERTING TRAINING INSTANCE NAME TO NOMENCLATURE
training_instance_name = convert_to_single_name(
solute=training_instance['solute'],
solvent=training_instance['cosolvent'],
mass_fraction=training_instance['mass_frac'],
temp=training_instance['temp'])
## DEFINING FULL TRAINING PATH
full_train_pickle_path = os.path.join(database_path,
training_instance_name)
## EXTRACTION PROTOCOL A PARTICULAR TRAINING EXAMPLE
training_data_for_instance = load_pickle(
full_train_pickle_path)
print(training_data_for_instance.shape
) # OUTPUT: (1001, 20, 20, 20, 3)
## I TAKE THE AVG, MANY OTHER WAYS ARE AVAILABLE
ensemble_avg1 = np.average(