apunkt/GPy
1
0
Fork 0
mirror of https://github.com/SheffieldML/GPy.git synced 2026-04-29 06:46:22 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

[missing data] general implementation for subsetting data

Browse source
This commit is contained in:
Max Zwiessele 2014-10-08 12:03:51 +01:00
parent d93fac8c13
commit fa7807ee6f
7 changed files with 329 additions and 108 deletions
Download patch file Download diff file Expand all files Collapse all files

59
GPy/testing/model_tests.py
View file

@ -20,7 +20,7 @@ class MiscTests(unittest.TestCase):
m = GPy.models.GPRegression(self.X, self.Y, kernel=k)
m.randomize()
m.likelihood.variance = .5
Kinv = np.linalg.pinv(k.K(self.X) + np.eye(self.N)*m.likelihood.variance)
Kinv = np.linalg.pinv(k.K(self.X) + np.eye(self.N) * m.likelihood.variance)
K_hat = k.K(self.X_new) - k.K(self.X_new, self.X).dot(Kinv).dot(k.K(self.X, self.X_new))
mu_hat = k.K(self.X_new, self.X).dot(Kinv).dot(m.Y_normalized)
@ -43,7 +43,7 @@ class MiscTests(unittest.TestCase):
Z = m.Z[:]
X = self.X[:]
#Not easy to check if woodbury_inv is correct in itself as it requires a large derivation and expression
# Not easy to check if woodbury_inv is correct in itself as it requires a large derivation and expression
Kinv = m.posterior.woodbury_inv
K_hat = k.K(self.X_new) - k.K(self.X_new, Z).dot(Kinv).dot(k.K(Z, self.X_new))
@ -51,13 +51,13 @@ class MiscTests(unittest.TestCase):
self.assertEquals(mu.shape, (self.N_new, self.D))
self.assertEquals(covar.shape, (self.N_new, self.N_new))
np.testing.assert_almost_equal(K_hat, covar)
#np.testing.assert_almost_equal(mu_hat, mu)
# np.testing.assert_almost_equal(mu_hat, mu)
mu, var = m._raw_predict(self.X_new)
self.assertEquals(mu.shape, (self.N_new, self.D))
self.assertEquals(var.shape, (self.N_new, 1))
np.testing.assert_almost_equal(np.diag(K_hat)[:, None], var)
#np.testing.assert_almost_equal(mu_hat, mu)
# np.testing.assert_almost_equal(mu_hat, mu)
def test_likelihood_replicate(self):
m = GPy.models.GPRegression(self.X, self.Y)
@ -110,6 +110,29 @@ class MiscTests(unittest.TestCase):
m2.kern.lengthscale = m.kern['.*lengthscale']
np.testing.assert_equal(m.log_likelihood(), m2.log_likelihood())
def test_missing_data(self):
from GPy import kern
from GPy.models import BayesianGPLVM
from GPy.examples.dimensionality_reduction import _simulate_matern
D1, D2, D3, N, num_inducing, Q = 13, 5, 8, 400, 3, 4
_, _, Ylist = _simulate_matern(D1, D2, D3, N, num_inducing, False)
Y = Ylist[0]
inan = np.random.binomial(1, .9, size=Y.shape).astype(bool) # 80% missing data
Ymissing = Y.copy()
Ymissing[inan] = np.nan
k = kern.Linear(Q, ARD=True) + kern.White(Q, np.exp(-2)) # + kern.bias(Q)
m = BayesianGPLVM(Ymissing, Q, init="random", num_inducing=num_inducing,
kernel=k, missing_data=True)
assert(m.checkgrad())
k = kern.RBF(Q, ARD=True) + kern.White(Q, np.exp(-2)) # + kern.bias(Q)
m = BayesianGPLVM(Ymissing, Q, init="random", num_inducing=num_inducing,
kernel=k, missing_data=True)
assert(m.checkgrad())
def test_likelihood_replicate_kern(self):
m = GPy.models.GPRegression(self.X, self.Y)
m2 = GPy.models.GPRegression(self.X, self.Y)
@ -158,7 +181,7 @@ class MiscTests(unittest.TestCase):
def test_model_optimize(self):
X = np.random.uniform(-3., 3., (20, 1))
Y = np.sin(X) + np.random.randn(20, 1) * 0.05
m = GPy.models.GPRegression(X,Y)
m = GPy.models.GPRegression(X, Y)
m.optimize()
print m
@ -219,8 +242,8 @@ class GradientTests(np.testing.TestCase):
mlp = GPy.kern.MLP(1)
self.check_model(mlp, model_type='GPRegression', dimension=1)
#TODO:
#def test_GPRegression_poly_1d(self):
# TODO:
# def test_GPRegression_poly_1d(self):
# ''' Testing the GP regression with polynomial kernel with white kernel on 1d data '''
# mlp = GPy.kern.Poly(1, degree=5)
# self.check_model(mlp, model_type='GPRegression', dimension=1)
@ -417,11 +440,11 @@ class GradientTests(np.testing.TestCase):
def test_gp_heteroscedastic_regression(self):
num_obs = 25
X = np.random.randint(0,140,num_obs)
X = X[:,None]
Y = 25. + np.sin(X/20.) * 2. + np.random.rand(num_obs)[:,None]
X = np.random.randint(0, 140, num_obs)
X = X[:, None]
Y = 25. + np.sin(X / 20.) * 2. + np.random.rand(num_obs)[:, None]
kern = GPy.kern.Bias(1) + GPy.kern.RBF(1)
m = GPy.models.GPHeteroscedasticRegression(X,Y,kern)
m = GPy.models.GPHeteroscedasticRegression(X, Y, kern)
self.assertTrue(m.checkgrad())
def test_gp_kronecker_gaussian(self):
@ -432,12 +455,12 @@ class GradientTests(np.testing.TestCase):
k1 = GPy.kern.RBF(1) # + GPy.kern.White(1)
k2 = GPy.kern.RBF(1) # + GPy.kern.White(1)
Y = np.random.randn(N1, N2)
Y = Y-Y.mean(0)
Y = Y/Y.std(0)
Y = Y - Y.mean(0)
Y = Y / Y.std(0)
m = GPy.models.GPKroneckerGaussianRegression(X1, X2, Y, k1, k2)
# build the model the dumb way
assert (N1*N2<1000), "too much data for standard GPs!"
assert (N1 * N2 < 1000), "too much data for standard GPs!"
yy, xx = np.meshgrid(X2, X1)
Xgrid = np.vstack((xx.flatten(order='F'), yy.flatten(order='F'))).T
kg = GPy.kern.RBF(1, active_dims=[0]) * GPy.kern.RBF(1, active_dims=[1])
@ -458,11 +481,11 @@ class GradientTests(np.testing.TestCase):
def test_gp_VGPC(self):
num_obs = 25
X = np.random.randint(0,140,num_obs)
X = X[:,None]
Y = 25. + np.sin(X/20.) * 2. + np.random.rand(num_obs)[:,None]
X = np.random.randint(0, 140, num_obs)
X = X[:, None]
Y = 25. + np.sin(X / 20.) * 2. + np.random.rand(num_obs)[:, None]
kern = GPy.kern.Bias(1) + GPy.kern.RBF(1)
m = GPy.models.GPVariationalGaussianApproximation(X,Y,kern)
m = GPy.models.GPVariationalGaussianApproximation(X, Y, kern)
self.assertTrue(m.checkgrad())