def test_kriging_f_order(n, m):
g = lk.RandomGenerator(123)
X = np.asarray(g.uniform(n, m), dtype=np.float64, order='F')
y = f(X)
rl = lk.WrappedPyKriging(y, X, "gauss")
y_pred, _stderr, _cov = rl.predict(X, True, True)
def test_kriging_c_order(n, m):
g = lk.RandomGenerator(123)
# X = np.asarray(g.uniform(n, m), dtype=np.float64, order='C')
X = g.uniform(n, m) # this is the default
y = f(X)
rl = lk.PyKriging(y, X, "gauss")
y_pred, _stderr, _cov = rl.predict(X, True, True)
def test_random():
g = lk.RandomGenerator(123)
expected_result = np.array([[0.71295532, 0.78002776],
[0.42847093, 0.41092437],
[0.69088485, 0.5796943],
[0.71915031, 0.13995076],
[0.49111893, 0.40101755]])
generated_result = g.uniform(5, 2)
# print(generated_result)
assert np.isclose(generated_result, expected_result).all()
def test_linear_regression_exact(n, m):
g = lk.RandomGenerator(123)
sol = np.asarray(g.uniform(m, 1), dtype=np.float64, order='F')
X = np.asarray(g.uniform(n, m), dtype=np.float64, order='F')
X[:, 0] = 1
y = X.dot(sol) # or X @ sol
rl = lk.LinearRegression()
rl.fit(y, X)
y2, _stderr = rl.predict(X)
eps = 1e-5
assert np.linalg.norm(y - y2, ord=np.inf) <= eps
def test_basic():
"""Test of manual binding"""
n = 40
m = 3
g = lk.RandomGenerator(123)
sol = np.asarray(g.uniform(m, 1), dtype=np.float64, order='F')
X = np.asarray(g.uniform(n, m), dtype=np.float64, order='F')
X[:, 0] = 1
y = X.dot(sol) # or X @ sol
rl = lk.PyLinearRegression()
rl.fit(y, X)
y2, _stderr = rl.predict(X)
eps = 1e-5
assert np.linalg.norm(y - y2, ord=2) <= eps
def test_linear_regression_noisy(n, m):
g = lk.RandomGenerator(123)
sol = np.asarray(g.uniform(m, 1), dtype=np.float64, order='F')
X = np.asarray(g.uniform(n, m), dtype=np.float64, order='F')
X[:, 0] = 1
y = X.dot(sol) # or X @ sol
e = 1e-8
noiser = lambda x: x * np.random.normal(1, e)
y = np.vectorize(noiser)(y)
# y = np.array([noiser(v) for v in y])
# y = np.array(list(map(noiser, y)))
rl = lk.LinearRegression()
rl.fit(y, X)
y2, _stderr = rl.predict(X)
eps = 1e-5
assert np.linalg.norm(y - y2, ord=np.inf) <= eps