def test_exercise_4(self):
try:
initial_w = np.array([[15], [-35.3]])
train_X_norm = standardize(self.train_X)
train_y_norm = standardize(self.train_y)
train_X_1 = add_feature_ones(train_X_norm)
learning_rate = 0.8
iterations = 20000
init = time.time()
w, weights_history, cost_history = batch_gradient_descent(
train_X_1, # noqa
train_y_norm, # noqa
initial_w, # noqa
learning_rate, # noqa
iterations) # noqa
init = time.time() - init
self.assertTrue(cost_history[-1] < cost_history[0], score=1 / 5)
self.assertTrue(type(w) == np.ndarray, score=1 / 5)
self.assertTrue(len(weights_history) == len(cost_history),
score=1 / 5) # noqa
self.assertTrue(init < 2.5, score=1 / 5)
test_X_norm = standardize(self.test_X)
test_X_1 = add_feature_ones(test_X_norm)
prediction = linear_regression_prediction(test_X_1, w)
prediction = (prediction * np.std(self.train_y)) + np.mean(
self.train_y) # noqa
r_2 = r_squared(self.test_y, prediction)
self.assertTrue(0.3 < r_2, score=1 / 5)
except NotImplementedError:
self.fail('Exercício 4) Falta fazer!')
def setUpClass(cls):
X, y = get_housing_prices_data(N=250, verbose=False)
train_X_norm = standardize(X)
train_y_norm = standardize(y)
cls.train_y_norm = train_y_norm.astype("float64")
cls.train_X_1 = add_feature_ones(train_X_norm)
cls.total_score = 0