def predict(self, input_data: InputData, verbose=False) -> OutputData:
if len(self.nodes_from) == 0:
raise ValueError('')
parent_results = []
if verbose:
print(f'Obtain predictions from all parent nodes: {self.model}')
for parent in self._nodes_from_with_fixed_order():
parent_results.append(parent.predict(input_data=input_data))
target = input_data.target
secondary_input = Data.from_predictions(outputs=parent_results,
target=target)
if verbose:
print(
f'Obtain prediction in secondary node with model: {self.model}'
)
preprocessed_data = copy(secondary_input)
preprocessed_data.features = self.cache.actual_cached_state.preprocessor.apply(
preprocessed_data.features)
evaluation_result = self.model.predict(
fitted_model=self.cache.actual_cached_state.model,
data=preprocessed_data)
return OutputData(idx=input_data.idx,
features=input_data.features,
predict=evaluation_result,
task_type=input_data.task_type)
def fit(self, input_data: InputData, verbose=False) -> OutputData:
if verbose:
print(f'Trying to fit primary node with model: {self.model}')
model_predict = self._fit_using_cache(input_data=input_data, verbose=verbose)
return OutputData(idx=input_data.idx,
features=input_data.features,
predict=model_predict, task=input_data.task,
data_type=self.model.output_datatype(input_data.data_type))
def output_dataset():
task_type = MachineLearningTasksEnum.classification
samples = 1000
x = 10.0 * np.random.rand(samples, ) - 5.0
x = np.expand_dims(x, axis=1)
threshold = 0.5
y = 1.0 / (1.0 + np.exp(np.power(x, -1.0)))
classes = np.array([0.0 if val <= threshold else 1.0 for val in y])
classes = np.expand_dims(classes, axis=1)
data = OutputData(idx=np.arange(0, 100),
features=x,
predict=classes,
task_type=task_type)
return data
def predict(self, input_data: InputData, verbose=False) -> OutputData:
if verbose:
print(f'Predict in primary node by model: {self.model}')
if not self.cache:
raise ValueError('Model must be fitted before predict')
preprocessed_data = transformation_function_for_data(input_data.data_type,
self.model.metadata.input_types)(input_data)
preprocessed_data.features = self.cache.actual_cached_state.preprocessor.apply(preprocessed_data.features)
predict_train = self.model.predict(fitted_model=self.cache.actual_cached_state.model,
data=preprocessed_data)
return OutputData(idx=input_data.idx,
features=input_data.features,
predict=predict_train, task=input_data.task,
data_type=self.model.output_datatype(input_data.data_type))
def predict(self, input_data: InputData, verbose=False) -> OutputData:
if len(self.nodes_from) == 0:
raise ValueError('')
parent_results = []
if verbose:
print(f'Obtain predictions from all parent nodes: {self.model}')
parent_nodes = self._nodes_from_with_fixed_order()
target = input_data.target
# TODO refactor
if any(['affects_target' in parent_node.model_tags for parent_node in parent_nodes]):
if len(parent_nodes) == 1:
# is the previous model is the model that changes target
parent_result = parent_nodes[0].predict(input_data=input_data)
target = parent_result.predict
parent_results.append(parent_result)
else:
raise NotImplementedError()
else:
for parent in parent_nodes:
parent_results.append(parent.predict(input_data=input_data))
secondary_input = Data.from_predictions(outputs=parent_results,
target=target)
if verbose:
print(f'Obtain prediction in secondary node with model: {self.model}')
preprocessed_data = transformation_function_for_data(input_data.data_type,
self.model.metadata.input_types)(secondary_input)
evaluation_result = self.model.predict(fitted_model=self.cache.actual_cached_state.model,
data=preprocessed_data)
return OutputData(idx=input_data.idx,
features=input_data.features,
predict=evaluation_result,
data_type=self.model.output_datatype(input_data.data_type),
task=input_data.task)
def fit(self, input_data: InputData, verbose=False) -> OutputData:
if len(self.nodes_from) == 0:
raise ValueError()
parent_results = []
if verbose:
print(f'Fit all parent nodes in secondary node with model: {self.model}')
target = input_data.target
parent_nodes = self._nodes_from_with_fixed_order()
are_nodes_affect_target = ['affects_target' in parent_node.model_tags for parent_node in parent_nodes]
if any(are_nodes_affect_target):
if len(parent_nodes) == 1:
# is the previous model is the model that changes target
parent_result = parent_nodes[0].fit(input_data=input_data)
target = parent_result.predict
parent_results.append(parent_result)
else:
raise NotImplementedError()
else:
for parent in parent_nodes:
parent_results.append(parent.fit(input_data=input_data))
secondary_input = Data.from_predictions(outputs=parent_results,
target=target)
if verbose:
print(f'Trying to fit secondary node with model: {self.model}')
model_predict = self._fit_using_cache(input_data=secondary_input, with_preprocessing=False)
return OutputData(idx=input_data.idx,
features=input_data.features,
predict=model_predict,
task=input_data.task,
data_type=self.model.output_datatype(input_data.data_type))
def fit(self, input_data: InputData, verbose=False) -> OutputData:
if len(self.nodes_from) == 0:
raise ValueError()
parent_results = []
if verbose:
print(
f'Fit all parent nodes in secondary node with model: {self.model}'
)
for parent in self._nodes_from_with_fixed_order():
parent_results.append(parent.fit(input_data=input_data))
target = input_data.target
secondary_input = Data.from_predictions(outputs=parent_results,
target=target)
if verbose:
print(f'Trying to fit secondary node with model: {self.model}')
model_predict = self._fit_using_cache(input_data=secondary_input)
return OutputData(idx=input_data.idx,
features=input_data.features,
predict=model_predict,
task_type=input_data.task_type)
def output_from_prediction(self, input_data, prediction):
return OutputData(idx=input_data.idx,
features=input_data.features,
predict=prediction, task=input_data.task,
data_type=self.model.output_datatype(input_data.data_type))
def keras_model_predict(model, input_data: InputData):
evaluation_result = model.predict(input_data.features)
return OutputData(idx=input_data.idx,
features=input_data.features,
predict=evaluation_result,
task_type=input_data.task_type)