def test_predict_03(self):
np.random.seed(1)
x, y = get_xy()
gpr = GPR()
gpr.fit(x, y, optimization_restart=50)
test_y = gpr.predict(x)
np.testing.assert_array_almost_equal(y, test_y, decimal=6)
def test_optimal_mu(self):
x, y = get_xy()
gpr = GPR()
gpr.fit(x, y, optimization_restart=10)
new_mu = gpr.optimal_mu(np.array([[-1, 1], [-1, 1], [-1, 1], [-1, 1]]))
np.testing.assert_array_almost_equal(np.abs(new_mu),
np.ones(4).reshape(1, -1))
def test_predict_02(self):
np.random.seed(42)
x, y = get_xy()
gpr = GPR()
gpr.fit(x, y, optimization_restart=50)
test_y, variance = gpr.predict(x[:4], return_variance=True)
true_var = np.array([[0.0242762038, 0.0029760019],
[0.0244903294, 0.0030022514],
[0.0247701249, 0.0030365513],
[0.0232064664, 0.0028448636]])
np.testing.assert_array_almost_equal(true_var, variance, decimal=6)
def test_predict_02(self):
np.random.seed(42)
x, y = get_xy()
gpr = GPR()
gpr.fit(x, y, optimization_restart=50)
test_y, variance = gpr.predict(x[:4], return_variance=True)
true_var = np.array([[5.761689e-06, 2.017326e-06],
[5.761686e-06, 2.017325e-06],
[5.761692e-06, 2.017327e-06],
[5.761695e-06, 2.017328e-06]])
np.testing.assert_array_almost_equal(true_var, variance, decimal=6)
def test_predict_04(self):
pod = POD(method='svd', rank=3)
gpr = GPR()
db = Database(param, snapshots.T)
rom = ROM(db, pod, gpr).fit()
pred_sol = rom.predict(db.parameters)
assert pred_sol.shape == db.snapshots.shape
def test_predict_02(self):
np.random.seed(117)
pod = POD(method='svd', rank=4)
gpr = GPR()
db = Database(param, snapshots.T)
rom = ROM(db, pod, gpr).fit()
pred_sol = rom.predict([-.45, -.45])
np.testing.assert_allclose(pred_sol, pred_sol_gpr, rtol=1e-4, atol=1e-5)
def test_predict_03(self):
pod = POD(method='svd', rank=3)
gpr = GPR()
db = Database(param, snapshots.T)
#rom = ROM(db, pod, RBF()).fit()
#pred_sol = rom.predict([-.45, -.45])
#print(pred_sol)
rom = ROM(db, pod, gpr).fit()
pred_sol = rom.predict(db.parameters[2])
assert pred_sol.shape == db.snapshots[0].shape
def test_loo_error_02(self):
pod = POD()
gpr = GPR()
db = Database(param, snapshots.T)
rom = ROM(db, pod, gpr)
err = rom.loo_error(normalizer=False)
np.testing.assert_allclose(
err[0],
np.array(0.639247),
rtol=1e-3)
def test_kfold_cv_error_03(self):
pod = POD()
gpr = GPR()
db = Database(param, snapshots.T)
rom = ROM(db, pod, gpr)
err = rom.kfold_cv_error(n_splits=3, normalizer=False)
np.testing.assert_allclose(
err,
np.array([0.664149, 1.355502, 0.379874]),
rtol=1e-3)
def test_predict_01(self):
x, y = get_xy()
gpr = GPR()
gpr.fit(x, y, optimization_restart=50)
test_y, variance = gpr.predict(x, return_variance=True)
np.testing.assert_array_almost_equal(y, test_y, decimal=6)
def test_types(self):
x, y = get_xy()
gpr = GPR()
gpr.fit(x[:, 0], y[:, 0])
assert isinstance(gpr.model,
sklearn.gaussian_process.GaussianProcessRegressor)
def test_constructor_empty(self):
gpr = GPR()
def test_fit(self):
x, y = get_xy()
gpr = GPR()
gpr.fit(x, y)
assert isinstance(gpr.model, GPy.models.GPRegression)
def test_fit_mono(self):
x, y = get_xy()
gpr = GPR()
gpr.fit(x[:, 0], y[:, 0])
assert isinstance(gpr.model, GPy.models.GPRegression)
#
# Moreover, new state-of-the-art methods will arrive, so we invite you to read the [documentation](https://mathlab.github.io/EZyRB/) for the complete list of all the possibilities!
#
# In the next cell, we create two dictionaries with the objects, such that we can easily test everything with simple `for` cycles. **WARNING** since several methods require the solution of an optimization problem (eg. GPR, ANN, AE), the cell may require some minutes to be run.
# In[9]:
reductions = {
'POD': POD('svd', rank=10),
'AE': AE([200, 100, 10], [10, 100, 200], nn.Tanh(), nn.Tanh(), 10),
}
approximations = {
# 'Linear': Linear(),
'RBF': RBF(),
'GPR': GPR(),
'KNeighbors': KNeighborsRegressor(),
'RadiusNeighbors': RadiusNeighborsRegressor(),
'ANN': ANN([20, 20], nn.Tanh(), 10),
}
header = '{:10s}'.format('')
for name in approximations:
header += ' {:>15s}'.format(name)
print(header)
for redname, redclass in reductions.items():
row = '{:10s}'.format(redname)
for approxname, approxclass in approximations.items():
rom = ROM(db, redclass, approxclass)
rom.fit()