def test_dot_1D_1D(self):
Xn = numpy.array(self.x_array[0])
Yn = numpy.array(self.y_array[0])
Xmy = my_np.array(self.x_array[0])
Ymy = my_np.array(self.y_array[0])
original = numpy.dot(Xn.T, Yn)
my = my_np.dot(Xmy.T, Ymy)
self.assertEqual(original, my)
def test_exp_list(self):
nb_n = numpy.array(self.x_array[0])
nb_my = my_np.array(self.x_array[0])
original = numpy.exp(nb_n)
my = my_np.exp(nb_my)
original = [round(nb, 12) for nb in original.tolist()]
my = [round(nb, 12) for nb in list(my)]
self.assertEqual(original, my)
def fit(self, X, Y):
"""Calculate optimal thetas to predict future data"""
def one_vs_all(house):
return [1 if y == house else 0 for y in Y]
self.theta = []
X = np.insert(X, 0, 1, axis=1)
m = Courses.m
theta_nb = len(X[0])
for house in range(4):
y_ova = one_vs_all(house)
theta = np.zeros(theta_nb)
for _ in range(self.iterations):
theta -= self.gradient_descent(X, theta, y_ova, m)
self.theta.append(theta)
def gradient_descent(self, X, theta, Y, m):
"""Update theta values through vectorized implementation of GD"""
Z = X.dot(theta)
H = Predict.g(Z)
gradient = np.dot(X.T, (H - Y)) / m
return self.alpha * gradient
def test_rmath(self, nb=3):
original = [f(nb, self.Xn).tolist() for f in self.math]
my = [list(f(nb, self.Xmy)) for f in self.math]
self.assertEqual(original, my)
Xn = numpy.array(self.x_array[0])
Xmy = my_np.array(self.x_array[0])
Y = self.y_array[0]
original = [f(Y, Xn).tolist() for f in self.math]
my = [list(f(Y, Xmy)) for f in self.math]
original.append((Y**Xn).tolist())
my.append(list(Y**Xmy))
self.assertEqual(original, my)
def test_math(self, nb=3):
original = [f(self.Xn, nb).tolist() for f in self.math]
my = [list(f(self.Xmy, nb)) for f in self.math]
self.assertEqual(original, my)
Xn = numpy.array(self.x_array[0])
Xmy = my_np.array(self.x_array[0])
Y = self.y_array[0]
original = [f(Xn, Y).tolist() for f in self.math]
my = [list(f(Xmy, Y)) for f in self.math]
self.assertEqual(original, my)
Xn -= nb
Xmy -= nb
self.assertEqual(Xn.tolist(), list(Xmy))
self.assertEqual((-Xn).tolist(), list(-Xmy))
class Tests(unittest.TestCase):
x_array = [[1, 2, 3], [1, 2, 3]]
y_array = [[7, 8, 9], [7, 8, 9]]
Xn = numpy.array(x_array)
Yn = numpy.array(y_array)
Xmy = my_np.array(x_array)
Ymy = my_np.array(y_array)
def test_arrays(self):
self.assertEqual(self.Xn.tolist(), list(self.Xmy))
self.assertEqual(self.Yn.tolist(), list(self.Ymy))
def test_astype(self, dtype=float):
self.assertEqual(
self.Xn.astype(dtype).tolist(), list(self.Xmy.astype(dtype)))
def test_exp_int(self, nb=3):
original = numpy.exp(nb)
my = my_np.exp(nb)
self.assertEqual(round(original, 12), round(my, 12))
def test_exp_list(self):
nb_n = numpy.array(self.x_array[0])
nb_my = my_np.array(self.x_array[0])
original = numpy.exp(nb_n)
my = my_np.exp(nb_my)
original = [round(nb, 12) for nb in original.tolist()]
my = [round(nb, 12) for nb in list(my)]
self.assertEqual(original, my)
def test_dot_2D_2D(self):
original = numpy.dot(self.Xn.T, self.Yn)
my = my_np.dot(self.Xmy.T, self.Ymy)
self.assertEqual(original.tolist(), list(my))
def test_dot_2D_1D(self, y=[1, 2]):
original = numpy.dot(self.Xn.T, y)
my = my_np.dot(self.Xmy.T, y)
self.assertEqual(original.tolist(), list(my))
def test_dot_1D_1D(self):
Xn = numpy.array(self.x_array[0])
Yn = numpy.array(self.y_array[0])
Xmy = my_np.array(self.x_array[0])
Ymy = my_np.array(self.y_array[0])
original = numpy.dot(Xn.T, Yn)
my = my_np.dot(Xmy.T, Ymy)
self.assertEqual(original, my)
def test_array_dot_2D_2D(self):
original = self.Xn.dot(self.Yn.T)
my = self.Xmy.dot(self.Ymy.T)
self.assertEqual(original.tolist(), list(my))
math = [
lambda a, b: a + b,
lambda a, b: a - b,
lambda a, b: a * b,
lambda a, b: a / b,
]
def test_math(self, nb=3):
original = [f(self.Xn, nb).tolist() for f in self.math]
my = [list(f(self.Xmy, nb)) for f in self.math]
self.assertEqual(original, my)
Xn = numpy.array(self.x_array[0])
Xmy = my_np.array(self.x_array[0])
Y = self.y_array[0]
original = [f(Xn, Y).tolist() for f in self.math]
my = [list(f(Xmy, Y)) for f in self.math]
self.assertEqual(original, my)
Xn -= nb
Xmy -= nb
self.assertEqual(Xn.tolist(), list(Xmy))
self.assertEqual((-Xn).tolist(), list(-Xmy))
def test_rmath(self, nb=3):
original = [f(nb, self.Xn).tolist() for f in self.math]
my = [list(f(nb, self.Xmy)) for f in self.math]
self.assertEqual(original, my)
Xn = numpy.array(self.x_array[0])
Xmy = my_np.array(self.x_array[0])
Y = self.y_array[0]
original = [f(Y, Xn).tolist() for f in self.math]
my = [list(f(Y, Xmy)) for f in self.math]
original.append((Y**Xn).tolist())
my.append(list(Y**Xmy))
self.assertEqual(original, my)
def test_insert_numpy_array(self):
original = numpy.insert(self.Xn, 0, 1, axis=1)
my = my_np.insert(self.Xmy, 0, 1, axis=1)
self.assertEqual(original.tolist(), list(my))
def test_insert_list(self, sample=[[1, 2, 3], [1, 2, 3]]):
original = numpy.insert(sample, 0, 1, axis=1)
my = my_np.insert(sample, 0, 1, axis=1)
self.assertEqual(original.tolist(), list(my))
def test_transpose(self):
self.assertEqual(self.Xn.T.tolist(), list(self.Xmy.T))
self.assertEqual(self.Yn.T.tolist(), list(self.Ymy.T))
def test_zeros(self, nb=5):
self.assertEqual(numpy.zeros(nb).tolist(), list(my_np.zeros(nb)))
def test_dot_2D_1D(self, y=[1, 2]):
original = numpy.dot(self.Xn.T, y)
my = my_np.dot(self.Xmy.T, y)
self.assertEqual(original.tolist(), list(my))
def test_dot_2D_2D(self):
original = numpy.dot(self.Xn.T, self.Yn)
my = my_np.dot(self.Xmy.T, self.Ymy)
self.assertEqual(original.tolist(), list(my))
def test_exp_int(self, nb=3):
original = numpy.exp(nb)
my = my_np.exp(nb)
self.assertEqual(round(original, 12), round(my, 12))
def test_zeros(self, nb=5):
self.assertEqual(numpy.zeros(nb).tolist(), list(my_np.zeros(nb)))
def test_insert_list(self, sample=[[1, 2, 3], [1, 2, 3]]):
original = numpy.insert(sample, 0, 1, axis=1)
my = my_np.insert(sample, 0, 1, axis=1)
self.assertEqual(original.tolist(), list(my))
def test_insert_numpy_array(self):
original = numpy.insert(self.Xn, 0, 1, axis=1)
my = my_np.insert(self.Xmy, 0, 1, axis=1)
self.assertEqual(original.tolist(), list(my))