def eval_with_existing(self, x_train, y_train, x_val, y_val, x_test,
y_test, weight_file):
model = self.load_model(weight_file)
y_pred_on_train = reshape_data_into_2_dim(model.predict(x_train))
y_pred_proba_on_train = model.predict_proba(x_train)
if x_val is not None:
y_pred_on_val = reshape_data_into_2_dim(model.predict(x_val))
y_pred_proba_on_val = model.predict_proba(x_val)
else:
y_pred_on_val, y_pred_proba_on_val = None, None
if x_test is not None:
y_pred_on_test = reshape_data_into_2_dim(model.predict(x_test))
y_pred_proba_on_test = model.predict_proba(x_test)
else:
y_pred_on_test, y_pred_proba_on_test = None, None
output_classification_result(
y_train=y_train,
y_pred_on_train=y_pred_on_train,
y_pred_proba_on_train=y_pred_proba_on_train,
y_val=y_val,
y_pred_on_val=y_pred_on_val,
y_pred_proba_on_val=y_pred_proba_on_val,
y_test=y_test,
y_pred_on_test=y_pred_on_test,
y_pred_proba_on_test=y_pred_proba_on_test)
return
def train_and_predict(self, x_train, y_train, x_test, y_test, weight_file):
model = self.setup_model()
model.fit(x_train, y_train)
y_pred_on_train = reshape_data_into_2_dim(model.predict(x_train))
if x_test is not None:
y_pred_on_test = reshape_data_into_2_dim(model.predict(x_test))
else:
y_pred_on_test = None
output_classification_result(y_train=y_train,
y_pred_on_train=y_pred_on_train,
y_val=None,
y_pred_on_val=None,
y_test=y_test,
y_pred_on_test=y_pred_on_test)
np.savez('output_{}/num_class_{}_index_{}'.format(
mode, number_of_class, index),
y_train=y_train,
y_pred_on_train=y_pred_on_train,
y_test=y_test,
y_pred_on_test=y_pred_on_test)
self.save_model(model, weight_file)
return
def predict_with_existing(self, X_data, weight_file):
model = self.load_model(weight_file)
y_pred = reshape_data_into_2_dim(model.predict_proba(X_data)[:, 1])
return y_pred