mirror of
https://github.com/SheffieldML/GPy.git
synced 2026-05-18 13:55:14 +02:00
[basisfunc] kernel tests and model tests
This commit is contained in:
Max Zwiessele
parent
c3c98652d1
commit
ba48c00e9d
5 changed files with 119 additions and 19 deletions
|
|
@ -477,6 +477,34 @@ class KernelGradientTestsContinuous(unittest.TestCase):
|
|||
k.randomize()
|
||||
self.assertTrue(check_kernel_gradient_functions(k, X=X, X2=X2, verbose=verbose, fixed_X_dims=[0]))
|
||||
|
||||
def test_basis_func_linear_slope(self):
|
||||
start_stop = np.random.uniform(self.X.min(0), self.X.max(0), (4, self.X.shape[1])).T
|
||||
start_stop.sort(axis=1)
|
||||
ks = []
|
||||
for i in range(start_stop.shape[0]):
|
||||
start, stop = np.split(start_stop[i], 2)
|
||||
ks.append(GPy.kern.LinearSlopeBasisFuncKernel(1, start, stop, ARD=i%2==0, active_dims=[i]))
|
||||
k = GPy.kern.Add(ks)
|
||||
self.assertTrue(check_kernel_gradient_functions(k, X=self.X, X2=self.X2, verbose=verbose))
|
||||
|
||||
def test_basis_func_changepoint(self):
|
||||
points = np.random.uniform(self.X.min(0), self.X.max(0), (self.X.shape[1]))
|
||||
ks = []
|
||||
for i in range(points.shape[0]):
|
||||
ks.append(GPy.kern.ChangePointBasisFuncKernel(1, points[i], ARD=i%2==0, active_dims=[i]))
|
||||
k = GPy.kern.Add(ks)
|
||||
self.assertTrue(check_kernel_gradient_functions(k, X=self.X, X2=self.X2, verbose=verbose))
|
||||
|
||||
def test_basis_func_domain(self):
|
||||
start_stop = np.random.uniform(self.X.min(0), self.X.max(0), (4, self.X.shape[1])).T
|
||||
start_stop.sort(axis=1)
|
||||
ks = []
|
||||
for i in range(start_stop.shape[0]):
|
||||
start, stop = np.split(start_stop[i], 2)
|
||||
ks.append(GPy.kern.DomainKernel(1, start, stop, ARD=i%2==0, active_dims=[i]))
|
||||
k = GPy.kern.Add(ks)
|
||||
self.assertTrue(check_kernel_gradient_functions(k, X=self.X, X2=self.X2, verbose=verbose))
|
||||
|
||||
class KernelTestsMiscellaneous(unittest.TestCase):
|
||||
def setUp(self):
|
||||
N, D = 100, 10
|
||||
|
|
|
|||
|
|
@ -91,23 +91,23 @@ class MiscTests(unittest.TestCase):
|
|||
k = GPy.kern.RBF(1)
|
||||
m2 = GPy.models.GPRegression(self.X, (Y-mu)/std, kernel=k, normalizer=False)
|
||||
m2[:] = m[:]
|
||||
|
||||
|
||||
mu1, var1 = m.predict(m.X, full_cov=True)
|
||||
mu2, var2 = m2.predict(m2.X, full_cov=True)
|
||||
np.testing.assert_allclose(mu1, (mu2*std)+mu)
|
||||
np.testing.assert_allclose(var1, var2*std**2)
|
||||
|
||||
|
||||
mu1, var1 = m.predict(m.X, full_cov=False)
|
||||
mu2, var2 = m2.predict(m2.X, full_cov=False)
|
||||
|
||||
|
||||
np.testing.assert_allclose(mu1, (mu2*std)+mu)
|
||||
np.testing.assert_allclose(var1, var2*std**2)
|
||||
|
||||
q50n = m.predict_quantiles(m.X, (50,))
|
||||
q50 = m2.predict_quantiles(m2.X, (50,))
|
||||
|
||||
|
||||
np.testing.assert_allclose(q50n[0], (q50[0]*std)+mu)
|
||||
|
||||
|
||||
# Test variance component:
|
||||
qs = np.array([2.5, 97.5])
|
||||
# The quantiles get computed before unormalization
|
||||
|
|
@ -181,8 +181,8 @@ class MiscTests(unittest.TestCase):
|
|||
Y_mu_true = 2*X_pred_mu
|
||||
Y_var_true = 4*X_pred_var
|
||||
Y_mu_pred, Y_var_pred = m.predict_noiseless(X_pred)
|
||||
np.testing.assert_allclose(Y_mu_true, Y_mu_pred, rtol=1e-4)
|
||||
np.testing.assert_allclose(Y_var_true, Y_var_pred, rtol=1e-4)
|
||||
np.testing.assert_allclose(Y_mu_true, Y_mu_pred, rtol=1e-3)
|
||||
np.testing.assert_allclose(Y_var_true, Y_var_pred, rtol=1e-3)
|
||||
|
||||
def test_sparse_raw_predict(self):
|
||||
k = GPy.kern.RBF(1)
|
||||
|
|
@ -376,18 +376,18 @@ class MiscTests(unittest.TestCase):
|
|||
|
||||
warp_k = GPy.kern.RBF(1)
|
||||
warp_f = GPy.util.warping_functions.IdentityFunction(closed_inverse=False)
|
||||
warp_m = GPy.models.WarpedGP(self.X, self.Y, kernel=warp_k,
|
||||
warp_m = GPy.models.WarpedGP(self.X, self.Y, kernel=warp_k,
|
||||
warping_function=warp_f)
|
||||
warp_m.optimize()
|
||||
warp_preds = warp_m.predict(self.X)
|
||||
|
||||
warp_k_exact = GPy.kern.RBF(1)
|
||||
warp_f_exact = GPy.util.warping_functions.IdentityFunction()
|
||||
warp_m_exact = GPy.models.WarpedGP(self.X, self.Y, kernel=warp_k_exact,
|
||||
warp_m_exact = GPy.models.WarpedGP(self.X, self.Y, kernel=warp_k_exact,
|
||||
warping_function=warp_f_exact)
|
||||
warp_m_exact.optimize()
|
||||
warp_preds_exact = warp_m_exact.predict(self.X)
|
||||
|
||||
|
||||
np.testing.assert_almost_equal(preds, warp_preds, decimal=4)
|
||||
np.testing.assert_almost_equal(preds, warp_preds_exact, decimal=4)
|
||||
|
||||
|
|
@ -406,18 +406,18 @@ class MiscTests(unittest.TestCase):
|
|||
|
||||
warp_k = GPy.kern.RBF(1)
|
||||
warp_f = GPy.util.warping_functions.LogFunction(closed_inverse=False)
|
||||
warp_m = GPy.models.WarpedGP(self.X, Y, kernel=warp_k,
|
||||
warp_m = GPy.models.WarpedGP(self.X, Y, kernel=warp_k,
|
||||
warping_function=warp_f)
|
||||
warp_m.optimize()
|
||||
warp_preds = warp_m.predict(self.X, median=True)[0]
|
||||
|
||||
warp_k_exact = GPy.kern.RBF(1)
|
||||
warp_f_exact = GPy.util.warping_functions.LogFunction()
|
||||
warp_m_exact = GPy.models.WarpedGP(self.X, Y, kernel=warp_k_exact,
|
||||
warp_m_exact = GPy.models.WarpedGP(self.X, Y, kernel=warp_k_exact,
|
||||
warping_function=warp_f_exact)
|
||||
warp_m_exact.optimize(messages=True)
|
||||
warp_preds_exact = warp_m_exact.predict(self.X, median=True)[0]
|
||||
|
||||
|
||||
np.testing.assert_almost_equal(np.exp(preds), warp_preds, decimal=4)
|
||||
np.testing.assert_almost_equal(np.exp(preds), warp_preds_exact, decimal=4)
|
||||
|
||||
|
|
@ -435,7 +435,7 @@ class MiscTests(unittest.TestCase):
|
|||
|
||||
warp_m = GPy.models.WarpedGP(X, Y)#, kernel=warp_k)#, warping_function=warp_f)
|
||||
warp_m['.*\.d'].constrain_fixed(1.0)
|
||||
warp_m.optimize_restarts(parallel=False, robust=False, num_restarts=5,
|
||||
warp_m.optimize_restarts(parallel=False, robust=False, num_restarts=5,
|
||||
max_iters=max_iters)
|
||||
warp_m.predict(X)
|
||||
warp_m.predict_quantiles(X)
|
||||
|
|
@ -444,7 +444,7 @@ class MiscTests(unittest.TestCase):
|
|||
warp_m.plot()
|
||||
warp_m.predict_in_warped_space = True
|
||||
warp_m.plot()
|
||||
|
||||
|
||||
def test_offset_regression(self):
|
||||
#Tests GPy.models.GPOffsetRegression. Using two small time series
|
||||
#from a sine wave, we confirm the algorithm determines that the
|
||||
|
|
@ -465,6 +465,53 @@ class MiscTests(unittest.TestCase):
|
|||
m.optimize()
|
||||
assert np.abs(m.offset[0]-offset)<0.1, ("GPOffsetRegression model failing to estimate correct offset (value estimated = %0.2f instead of %0.2f)" % (m.offset[0], offset))
|
||||
|
||||
def test_logistic_basis_func_gradients(self):
|
||||
X = np.random.uniform(-4, 4, (20, 5))
|
||||
points = np.random.uniform(X.min(0), X.max(0), X.shape[1])
|
||||
ks = []
|
||||
for i in range(points.shape[0]):
|
||||
if (i%2==0) and (i%3!=0):
|
||||
self.assertRaises(AssertionError, GPy.kern.LogisticBasisFuncKernel, 1, points, ARD=i%2==0, ARD_slope=i%3==0, active_dims=[i])
|
||||
else:
|
||||
ks.append(GPy.kern.LogisticBasisFuncKernel(1, points, ARD=i%2==0, ARD_slope=i%3==0, active_dims=[i]))
|
||||
k = GPy.kern.Add(ks)
|
||||
k.randomize()
|
||||
|
||||
Y = np.random.normal(0, 1, (X.shape[0], 1))
|
||||
m = GPy.models.GPRegression(X, Y, kernel=k.copy())
|
||||
assert m.checkgrad()
|
||||
|
||||
def test_posterior_inf_basis_funcs(self):
|
||||
X = np.random.uniform(-4, 1, (50, 1))
|
||||
|
||||
# Logistic:
|
||||
k = GPy.kern.LogisticBasisFuncKernel(1, [0, -2])
|
||||
|
||||
true_w = [1, 2]
|
||||
true_slope = [5, -2]
|
||||
|
||||
Y = 0
|
||||
for w, s, c in zip(true_w, true_slope, k.centers[0]):
|
||||
Y += w/(1+np.exp(-s*(X-c)))
|
||||
Y += np.random.normal(0, .000001)
|
||||
|
||||
m = GPy.models.GPRegression(X,Y,kernel=k.copy())
|
||||
#m.likelihood.fix(1e-6)
|
||||
m.optimize()
|
||||
|
||||
wu, wv = m.kern.posterior_inf()
|
||||
#_sort = np.argsort(wu.flat)
|
||||
|
||||
#from scipy.stats import norm
|
||||
#confidence_intervals = np.array(norm.interval(.95, loc=wu.flat[_sort], scale=np.sqrt(np.diag(wv))[_sort])).T
|
||||
#for i in range(wu.size):
|
||||
# s,t = confidence_intervals[i]
|
||||
# v = true_w[i]
|
||||
# assert ((s<v)&(v<t)), "didnt find true w within the 95% confidence interval of the predicted values"
|
||||
|
||||
np.testing.assert_allclose(np.sort(wu.flat), np.sort(true_w), rtol=1e-4)
|
||||
np.testing.assert_allclose(np.diag(wv), 0, atol=1e-4)
|
||||
np.testing.assert_allclose(np.sort(m.kern.slope.flat), np.sort(true_slope), rtol=1e-4)
|
||||
|
||||
class GradientTests(np.testing.TestCase):
|
||||
def setUp(self):
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue