apunkt/GPy
1
0
Fork 0
mirror of https://github.com/SheffieldML/GPy.git synced 2026-05-10 20:42:39 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

stuff

Browse source
This commit is contained in:
beckdaniel 2016-01-14 15:28:48 +00:00
parent 4c843adc4c
commit 24e9d68a19
3 changed files with 10 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

2
GPy/models/warped_gp.py
View file

@ -160,7 +160,7 @@ class WarpedGP(GP):
mu_star, var_star = self._raw_predict(x_test)
fy = self.warping_function.f(y_test)
ll_lpd = self.likelihood.log_predictive_density(fy, mu_star, var_star, Y_metadata=Y_metadata)
return ll_lpd * self.warping_function.fgrad_y(y_test)
return ll_lpd - np.log(self.warping_function.fgrad_y(y_test))
if __name__ == '__main__':

4
GPy/testing/model_tests.py
View file

@ -307,7 +307,7 @@ class MiscTests(unittest.TestCase):
np.testing.assert_almost_equal(preds, warp_preds)
@unittest.skip('Comment this to plot the modified sine function')
#@unittest.skip('Comment this to plot the modified sine function')
def test_warped_gp_sine(self):
"""
A test replicating the sine regression problem from
@ -321,7 +321,6 @@ class MiscTests(unittest.TestCase):
warp_k = GPy.kern.RBF(1)
warp_f = GPy.util.warping_functions.TanhWarpingFunction_d(n_terms=2)
warp_m = GPy.models.WarpedGP(X[:, None], Y[:, None], kernel=warp_k, warping_function=warp_f)
warp_m['.*noise.variance.*'].constrain_fixed(0.1)
m = GPy.models.GPRegression(X[:, None], Y[:, None])
m.optimize_restarts(parallel=False, robust=True, num_restarts=5)
@ -330,6 +329,7 @@ class MiscTests(unittest.TestCase):
print(warp_m['.*warp.*'])
warp_m.predict_in_warped_space = False
warp_m.plot()
import ipdb; ipdb.set_trace()
warp_m.predict_in_warped_space = True
warp_m.plot()
m.plot()

9
GPy/util/warping_functions.py
View file

@ -152,7 +152,7 @@ class TanhWarpingFunction(WarpingFunction):
class TanhWarpingFunction_d(WarpingFunction):
def __init__(self, n_terms=3):
def __init__(self, n_terms=3, initial_y=None):
"""n_terms specifies the number of tanh terms to be used"""
self.n_terms = n_terms
self.num_parameters = 3 * self.n_terms + 1
@ -165,6 +165,7 @@ class TanhWarpingFunction_d(WarpingFunction):
self.d = Param('%s' % ('d'), 1.0, Logexp())
self.link_parameter(self.psi)
self.link_parameter(self.d)
self.initial_y = initial_y
def f(self, y):
"""
@ -187,7 +188,7 @@ class TanhWarpingFunction_d(WarpingFunction):
z += a*np.tanh(b*(y+c))
return z
def f_inv(self, z, max_iterations=1000, y=None):
def f_inv(self, z, max_iterations=100, y=None):
"""
calculate the numerical inverse of f
@ -195,12 +196,15 @@ class TanhWarpingFunction_d(WarpingFunction):
"""
z = z.copy()
if y is None:
# The idea here is to initialize y with +1 where
# z is positive and -1 where it is negative.
# For negative z, Newton-Raphson diverges
# if we initialize y with a positive value (and vice-versa).
y = ((z > 0) * 1.) - (z <= 0)
if self.initial_y is not None:
y *= self.initial_y
it = 0
update = np.inf
@ -213,6 +217,7 @@ class TanhWarpingFunction_d(WarpingFunction):
it += 1
if it == max_iterations:
print("WARNING!!! Maximum number of iterations reached in f_inv ")
print("Sum of updates: %.4f" % np.sum(update))
return y
def fgrad_y(self, y, return_precalc=False):