def _fit(self):
seed = self.random_state.randint(1000000)
for outer_fold in range(self.outer_cv_folds):
# First perform the fit for the outer cross validation
outer_train_indices, outer_test_indices = \
get_CV_fold(self.X_train, self.Y_train, fold=outer_fold,
folds=self.outer_cv_folds, shuffle=True,
random_state=seed)
self.outer_indices[outer_fold] = ((outer_train_indices,
outer_test_indices))
model = self.model_class(self.configuration, self.random_state)
self.outer_models[outer_fold] = model
self._fit_and_suppress_warnings(self.outer_models[outer_fold],
self.X_train[outer_train_indices],
self.Y_train[outer_train_indices])
# Then perform the fit for the inner cross validation
for inner_fold in range(self.inner_cv_folds):
X_train = self.X_train[outer_train_indices]
Y_train = self.Y_train[outer_train_indices]
inner_train_indices, inner_test_indices = \
get_CV_fold(X_train, Y_train,
fold=inner_fold, folds=self.inner_cv_folds,
shuffle=True, random_state=seed)
inner_train_indices = outer_train_indices[inner_train_indices]
inner_test_indices = outer_train_indices[inner_test_indices]
X_train = self.X_train[inner_train_indices]
Y_train = self.Y_train[inner_train_indices]
self.inner_indices[outer_fold][inner_fold] = \
((inner_train_indices, inner_test_indices))
model = self.model_class(self.configuration, self.random_state)
model = self._fit_and_suppress_warnings(model, X_train, Y_train)
self.inner_models[outer_fold][inner_fold] = model
def _fit(self):
seed = self.random_state.randint(1000000)
for outer_fold in range(self.outer_cv_folds):
# First perform the fit for the outer cross validation
outer_train_indices, outer_test_indices = \
get_CV_fold(self.X_train, self.Y_train, fold=outer_fold,
folds=self.outer_cv_folds, shuffle=True,
random_state=seed)
self.outer_indices[outer_fold] = ((outer_train_indices,
outer_test_indices))
model = self.model_class(self.configuration, self.random_state)
self.outer_models[outer_fold] = model
self.outer_models[outer_fold].fit(
self.X_train[outer_train_indices],
self.Y_train[outer_train_indices])
# Then perform the fit for the inner cross validation
for inner_fold in range(self.inner_cv_folds):
X_train = self.X_train[outer_train_indices]
Y_train = self.Y_train[outer_train_indices]
inner_train_indices, inner_test_indices = \
get_CV_fold(X_train, Y_train,
fold=inner_fold, folds=self.inner_cv_folds,
shuffle=True, random_state=seed)
inner_train_indices = outer_train_indices[inner_train_indices]
inner_test_indices = outer_train_indices[inner_test_indices]
X_train = self.X_train[inner_train_indices]
Y_train = self.Y_train[inner_train_indices]
self.inner_indices[outer_fold][inner_fold] = \
((inner_train_indices, inner_test_indices))
model = self.model_class(self.configuration, self.random_state)
model = model.fit(X_train, Y_train)
self.inner_models[outer_fold][inner_fold] = model
def get_train_test_split(self, fold):
return get_CV_fold(self.X_train,
self.Y_train,
fold=fold,
folds=self.cv_folds,
shuffle=True,
random_state=self.seed)
def _partial_fit_and_predict(self, fold):
model = self.model_class(self.configuration, self.seed)
train_indices, test_indices = \
get_CV_fold(self.X_train, self.Y_train, fold=fold,
folds=self.cv_folds, shuffle=True,
random_state=self.seed)
self.indices[fold] = ((train_indices, test_indices))
model.fit(self.X_train[train_indices],
self.Y_train[train_indices])
train_indices, test_indices = self.indices[fold]
self.Y_targets[fold] = self.Y_train[test_indices]
opt_pred = self.predict_function(self.X_train[test_indices],
model, self.task_type,
self.Y_train[train_indices])
X_valid = self.X_valid.copy()
valid_pred = self.predict_function(X_valid, model,
self.task_type,
self.Y_train[train_indices])
X_test = self.X_test.copy()
test_pred = self.predict_function(X_test, model,
self.task_type,
self.Y_train[train_indices])
return opt_pred, valid_pred, test_pred
def _partial_fit_and_predict(self, fold):
model = self.model_class(self.configuration, self.seed)
train_indices, test_indices = \
get_CV_fold(self.X_train, self.Y_train, fold=fold,
folds=self.cv_folds, shuffle=True,
random_state=self.seed)
self.indices[fold] = ((train_indices, test_indices))
model.fit(self.X_train[train_indices], self.Y_train[train_indices])
train_indices, test_indices = self.indices[fold]
self.Y_targets[fold] = self.Y_train[test_indices]
opt_pred = self.predict_function(self.X_train[test_indices], model,
self.task_type,
self.Y_train[train_indices])
X_valid = self.X_valid.copy()
valid_pred = self.predict_function(X_valid, model, self.task_type,
self.Y_train[train_indices])
X_test = self.X_test.copy()
test_pred = self.predict_function(X_test, model, self.task_type,
self.Y_train[train_indices])
return opt_pred, valid_pred, test_pred
def partial_fit(self, fold):
model = self.model_class(self.configuration, self.seed)
train_indices, test_indices = \
get_CV_fold(self.X_train, self.Y_train, fold=fold,
folds=self.cv_folds, shuffle=True,
random_state=self.seed)
self.indices[fold] = ((train_indices, test_indices))
self.models[fold] = model
self.models[fold].fit(self.X_train[train_indices],
self.Y_train[train_indices])
def partial_fit(self, fold):
model = self.model_class(self.configuration, self.seed)
train_indices, test_indices = \
get_CV_fold(self.X_train, self.Y_train, fold=fold,
folds=self.cv_folds, shuffle=True,
random_state=self.seed)
self.indices[fold] = ((train_indices, test_indices))
self.models[fold] = model
self.models[fold].fit(self.X_train[train_indices],
self.Y_train[train_indices])
def get_train_test_split(self, fold):
return get_CV_fold(self.X_train, self.Y_train, fold=fold,
folds=self.cv_folds, shuffle=True,
random_state=self.seed)