def setLearningRateIndex(self, index):
"""
This updates the pointers in the learning rates
:param index: the index
:return:
"""
if index >= len(self.learning_rates):
index = len(self.learning_rates) -1
logging.warning('Trying to set the learning rate on an index greater than learnign rates available')
self.current_learning_rate_index = index
self.total_epochs = self.learning_rates[self.current_learning_rate_index][EPOCHS_INDEX]
self.current_learning_rate = self.learning_rates[self.current_learning_rate_index][LEARNING_RATE_INDEX]
def predicted_values(self, as_list=False, add_header=False):
"""
Get an array of dictionaries (unless as_list=True) for predicted values
:return: predicted_values
"""
ret = []
# foreach row in the result extract only the predicted values
for row in self.data_array:
# prepare the result, either a dict or a list
if as_list:
ret_row = []
else:
ret_row = {}
# append predicted values
for col in self.predicted_columns:
col_index = self.columns.index(col)
if as_list:
ret_row += [row[col_index]]
else:
ret_row[self._getOrigColum(col)] = row[col_index]
# append confidence
col_index = self.columns.index(KEY_CONFIDENCE)
if as_list: # add confidence if its a dictionary
ret_row += row[col_index]
else:
if len(row) < col_index:
logging.warning(
'Output is smaller than expected, see transaction_output_data.py'
)
ret_row[KEY_CONFIDENCE] = 0
else:
try:
ret_row[KEY_CONFIDENCE] = row[col_index]
except:
ret_row[KEY_CONFIDENCE] = 0
# append row to result
ret += [ret_row]
# if add_header and as_list True, add the header to the result
if as_list and add_header:
header = [
self._getOrigColum(col) for col in self.predicted_columns
] + [KEY_CONFIDENCE]
ret = [header] + ret
return ret