apunkt/GPy
1
0
Fork 0
mirror of https://github.com/SheffieldML/GPy.git synced 2026-06-14 15:25:15 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

Update rv_transformation_tests.py

Browse source
This commit is contained in:
Max Zwiessele 2015-09-12 19:30:05 +01:00
parent a7c46499eb
commit 9b5323a949
1 changed files with 15 additions and 18 deletions
Download patch file Download diff file Expand all files Collapse all files

33
GPy/testing/rv_transformation_tests.py
View file

@ -30,20 +30,19 @@ class RVTransformationTestCase(unittest.TestCase):
m.theta.set_prior(prior)
m.theta.unconstrain()
m.theta.constrain(trans)
theta_s = prior.rvs(1e5)
if kde:
# The PDF of the transformed variables
p_phi = lambda phi : np.exp(-m._objective_grads(phi)[0])
# To the empirical PDF of:
phi_s = trans.finv(theta_s)
# which is essentially a kernel density estimation
kde = st.gaussian_kde(phi_s)
# We will compare the PDF here:
phi = np.linspace(phi_s.min(), phi_s.max(), 100)
# The transformed PDF of phi should be this:
pdf_phi = np.array([p_phi(p) for p in phi])
# The following test cannot be very accurate
self.assertTrue(np.linalg.norm(pdf_phi - kde(phi)) / np.linalg.norm(kde(phi)) <= 1e-1)
# The PDF of the transformed variables
p_phi = lambda phi : np.exp(-m._objective_grads(phi)[0])
# To the empirical PDF of:
phi_s = trans.finv(theta_s)
# which is essentially a kernel density estimation
kde = st.gaussian_kde(phi_s)
# We will compare the PDF here:
phi = np.linspace(phi_s.min(), phi_s.max(), 100)
# The transformed PDF of phi should be this:
pdf_phi = np.array([p_phi(p) for p in phi])
# The following test cannot be very accurate
self.assertTrue(np.linalg.norm(pdf_phi - kde(phi)) / np.linalg.norm(kde(phi)) <= 1e-1)
theta_s = prior.rvs(1e6)
# UNCOMMENT TO SEE GRAPHICAL COMPARISON
#import matplotlib.pyplot as plt
#fig, ax = plt.subplots()
@ -56,10 +55,8 @@ class RVTransformationTestCase(unittest.TestCase):
#plt.show(block=True)
# END OF PLOT
# Check the gradients at a few random points
np.random.seed(5)
for i in range(5):
#m.theta = theta_s[i]
m.randomize()
for i in range(10):
m.theta = theta_s[i]
self.assertTrue(m.checkgrad(verbose=True))
def test_Logexp(self):