def test_sigmoid_function2(self):
X_1 = np.array([[1200000]])
X_2 = np.array([[-25000]])
X_3 = np.array([[4, 5, 6]])
X_4 = magic(3)
X_5 = np.eye(2)
answer_1 = np.array([[1]])
answer_2 = np.zeros((1, 1))
answer_3 = np.array([[ 0.9820, 0.9933, 0.9975]])
answer_4 = np.array([[0.9997, 0.7311, 0.9975],
[0.9526, 0.9933, 0.9991],
[0.9820, 0.9999, 0.8808]])
answer_5 = np.array([[0.7311, 0.5000],
[0.5000, 0.7311]])
np.testing.assert_array_equal(sigmoid_function(X_1),
answer_1)
np.testing.assert_array_equal(sigmoid_function(X_2),
answer_2)
np.testing.assert_array_almost_equal(sigmoid_function(X_3),
answer_3, decimal=3)
np.testing.assert_array_almost_equal(sigmoid_function(X_4),
answer_4, decimal=3)
np.testing.assert_array_almost_equal(sigmoid_function(X_5),
answer_5, decimal=3)
def test_sigmoid_gradient(self):
actual = sigmoid_gradient(np.vstack((np.array([[-1., - 2., - 3.]]), magic(3))))
expected =np.array([[1.9661e-001, 1.0499e-001, 4.5177e-002],
[3.3524e-004, 1.9661e-001, 2.4665e-003],
[4.5177e-002, 6.6481e-003, 9.1022e-004],
[1.7663e-002, 1.2338e-004, 1.0499e-001]])
np.testing.assert_almost_equal(actual, expected, decimal=3)
def test_cost_function1(self):
X = np.c_[np.ones((3, 1)), magic(3)]
y = np.array([[1, 0, 1]]).T
theta = np.array([[-2, -1, 1, 2]]).T
j, grad = cost_function(X, y, theta)
answer_j = 4.6832
answer_grad = np.array([[0.31722,
0.87232,
1.64812,
2.23787]]).T
np.testing.assert_almost_equal(j, answer_j, decimal=3)
np.testing.assert_array_almost_equal(grad, answer_grad, decimal=3)
def test_cost_function_reg1(self):
X = np.c_[np.ones((3, 1)), magic(3)]
y = np.array([[1, 0, 1]]).T
theta = np.array([[-2, -1, 1, 2]]).T
lambda_ = 3
j, grad = cost_function_reg(X, y, theta, lambda_)
answer_j = 7.6832
answer_grad = np.array([[0.31722,
-0.12768,
2.64812,
4.23787]]).T
np.testing.assert_almost_equal(j, answer_j, decimal=3)
np.testing.assert_array_almost_equal(grad, answer_grad, decimal=3)
def test_opt_one_vs_all(self):
X = np.r_[magic(3),
np.sin(np.arange(1, 4)).reshape(1, 3),
np.cos(np.arange(1, 4)).reshape(1, 3)]
y = np.array([[1, 2, 2, 1, 3]])
y = y.T
num_labels = 3
theta = np.ones((X.shape[1] + 1, 1)) * 0.1
actual = ex3.opt_one_vs_all(X, y, theta, num_labels)
expected = np.array([[-0.559478, 0.619220, -0.550361, -0.093502],
[-5.472920, -0.471565, 1.261046, 0.634767],
[0.068368, -0.375582, -1.652262, -1.410138]])
np.testing.assert_array_almost_equal(actual, expected, decimal=3)