def test_iris_dim4(self):
from mlstatpy.ml.voronoi import voronoi_estimation_from_lr
data = load_iris()
X, y = data.data[:, :4], data.target
clr = LogisticRegression(solver="liblinear")
clr.fit(X, y)
C = [1., 0.]
D = 3.
self.assertRaise(
lambda: voronoi_estimation_from_lr(clr.coef_, clr.intercept_, C,
None), ValueError)
self.assertRaise(
lambda: voronoi_estimation_from_lr(clr.coef_, clr.intercept_, C,
[D]), ValueError)
C = [1., 0., 0., 0.]
points = voronoi_estimation_from_lr(clr.coef_,
clr.intercept_,
C,
D,
qr=False)
self.assertEqual(points.shape, (3, 4))
points2 = voronoi_estimation_from_lr(clr.coef_,
clr.intercept_,
C,
D,
qr=True)
self.assertEqual(points2.shape, (3, 4))
self.assertEqualArray(points2, points2, decimal=5)
def test_square(self):
from mlstatpy.ml.voronoi import voronoi_estimation_from_lr
Xs = []
Ys = []
n = 20
for i in range(0, 4):
for j in range(0, 3):
x1 = numpy.random.rand(n) + i * 1.1
x2 = numpy.random.rand(n) + j * 1.1
Xs.append(numpy.vstack([x1, x2]).T)
Ys.extend([i * 3 + j] * n)
X = numpy.vstack(Xs)
Y = numpy.array(Ys)
clr = LogisticRegression(solver="liblinear")
clr.fit(X, Y)
points = voronoi_estimation_from_lr(clr.coef_,
clr.intercept_,
qr=True,
verbose=False)
self.assertEqual(points.shape, (12, 2))
self.assertGreater(points.ravel().min(), -35)
self.assertLesser(points.ravel().max(), 7.2)
def test_iris(self):
from mlstatpy.ml import voronoi_estimation_from_lr
data = load_iris()
X, y = data.data[:, :2], data.target
clr = LogisticRegression(solver="liblinear")
clr.fit(X, y)
C = [1., 0.]
D = 3.
self.assertRaise(
lambda: voronoi_estimation_from_lr(clr.coef_, clr.intercept_, C,
None), ValueError)
self.assertRaise(
lambda: voronoi_estimation_from_lr(clr.coef_, clr.intercept_, C,
[D]), TypeError)
std = StringIO()
with redirect_stdout(std):
points = voronoi_estimation_from_lr(clr.coef_,
clr.intercept_,
C,
D,
qr=False,
verbose=True)
self.assertEqual(points.shape, (3, 2))
expected_values = numpy.array([[3., 4.137], [5.044, 0.281],
[5.497, 0.184]])
self.assertEqualArray(expected_values, points, decimal=2)
points = voronoi_estimation_from_lr(clr.coef_,
clr.intercept_,
C,
D,
qr=True,
verbose=True)
self.assertEqual(points.shape, (3, 2))
expected_values = numpy.array([[3., 4.137], [5.044, 0.281],
[5.497, 0.184]])
self.assertEqualArray(expected_values, points, decimal=2)
std = std.getvalue()
self.assertIn('[voronoi_estimation_from_lr] iter=', std)
def test_hexa_scale(self):
from mlstatpy.ml.voronoi import voronoi_estimation_from_lr
n = 4
a = math.pi * 2 / 3
points = []
Ys = []
for i in range(n):
for j in range(n):
dil = ((i + 1)**2 + (j + 1)**2)**0.6
for _ in range(0, 20):
x = i + j * math.cos(a)
y = j * math.sin(a)
points.append([x * dil, y * dil])
Ys.append(i * n + j)
mi = 0.5
for r in [0.1, 0.3, mi]:
nb = 6 if r == mi else 12
for k2 in range(0, nb):
ang = math.pi * 2 / nb * k2 + math.pi / 6
x = i + j * math.cos(a) + r * math.cos(ang)
y = j * math.sin(a) + r * math.sin(ang)
points.append([x * dil, y * dil])
Ys.append(i * n + j)
X = numpy.array(points)
Y = numpy.array(Ys)
clr = LogisticRegression(solver="liblinear")
clr.fit(X, Y)
std = StringIO()
with redirect_stdout(std):
points = voronoi_estimation_from_lr(clr.coef_,
clr.intercept_,
qr=True,
verbose=True,
max_iter=20)
self.assertEqual(points.shape, (16, 2))
self.assertGreater(points.ravel().min(), -15)
self.assertLesser(points.ravel().max(), 16)
std = std.getvalue()
self.assertIn('del P', std)
self.assertIn('[voronoi_estimation_from_lr] iter', std)