apunkt/GPy
1
0
Fork 0
mirror of https://github.com/SheffieldML/GPy.git synced 2026-05-08 19:42:39 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

[SparseGP] added self.full_values

Browse source
This commit is contained in:
Max Zwiessele 2014-10-31 17:10:02 +00:00
parent bfe8110c87
commit 0080d4b044
1 changed files with 7 additions and 31 deletions
Download patch file Download diff file Expand all files Collapse all files

38
GPy/core/sparse_gp.py
View file

@ -176,30 +176,6 @@ class SparseGP(GP):
value_indices:
dictionary holding indices for the update in full_values.
if the key exists the update rule is:def df(x):
m.stochastics.do_stochastics()
grads = m._grads(x)
print '\r',
message = "Lik: {: 6.4E} Grad: {: 6.4E} Dim: {} Lik: {} Len: {!s}".format(float(m.log_likelihood()), np.einsum('i,i->', grads, grads), m.stochastics.d, float(m.likelihood.variance), " ".join(["{:3.2E}".format(l) for l in m.kern.lengthscale.values]))
print message,
return grads
def grad_stop(threshold):
def inner(args):
g = args['gradient']
return np.sqrt(np.einsum('i,i->',g,g)) < threshold
return inner
def maxiter_stop(maxiter):
def inner(args):
return args['n_iter'] == maxiter
return inner
def optimize(m, maxiter=1000):
#opt = climin.RmsProp(m.optimizer_array.copy(), df, 1e-6, decay=0.9, momentum=0.9, step_adapt=1e-7)
opt = climin.Adadelta(m.optimizer_array.copy(), df, 1e-2, decay=0.9)
ret = opt.minimize_until((grad_stop(.1), maxiter_stop(maxiter)))
print
return ret
full_values[key][value_indices[key]] += current_values[key]
"""
for key in current_values.keys():
@ -251,7 +227,7 @@ def optimize(m, maxiter=1000):
dL_dKmm = None
self._log_marginal_likelihood = 0
full_values = self._outer_init_full_values()
self.full_values = self._outer_init_full_values()
if self.posterior is None:
woodbury_inv = np.zeros((self.num_inducing, self.num_inducing, self.output_dim))
@ -281,7 +257,7 @@ def optimize(m, maxiter=1000):
Lm, dL_dKmm,
subset_indices=dict(outputs=d, samples=ninan))
self._inner_take_over_or_update(full_values, current_values, value_indices)
self._inner_take_over_or_update(self.full_values, current_values, value_indices)
self._inner_values_update(current_values)
Lm = posterior.K_chol
@ -295,7 +271,7 @@ def optimize(m, maxiter=1000):
if self.posterior is None:
self.posterior = Posterior(woodbury_inv=woodbury_inv, woodbury_vector=woodbury_vector,
K=posterior._K, mean=None, cov=None, K_chol=posterior.K_chol)
self._outer_values_update(full_values)
self._outer_values_update(self.full_values)
def _outer_loop_without_missing_data(self):
self._log_marginal_likelihood = 0
@ -309,7 +285,7 @@ def optimize(m, maxiter=1000):
d = self.stochastics.d
posterior, log_marginal_likelihood, \
grad_dict, current_values, _ = self._inner_parameters_changed(
grad_dict, self.full_values, _ = self._inner_parameters_changed(
self.kern, self.X,
self.Z, self.likelihood,
self.Y_normalized[:, d], self.Y_metadata)
@ -317,7 +293,7 @@ def optimize(m, maxiter=1000):
self._log_marginal_likelihood += log_marginal_likelihood
self._outer_values_update(current_values)
self._outer_values_update(self.full_values)
woodbury_inv[:, :, d] = posterior.woodbury_inv[:, :, None]
woodbury_vector[:, d] = posterior.woodbury_vector
@ -331,8 +307,8 @@ def optimize(m, maxiter=1000):
elif self.stochastics:
self._outer_loop_without_missing_data()
else:
self.posterior, self._log_marginal_likelihood, self.grad_dict, full_values, _ = self._inner_parameters_changed(self.kern, self.X, self.Z, self.likelihood, self.Y_normalized, self.Y_metadata)
self._outer_values_update(full_values)
self.posterior, self._log_marginal_likelihood, self.grad_dict, self.full_values, _ = self._inner_parameters_changed(self.kern, self.X, self.Z, self.likelihood, self.Y_normalized, self.Y_metadata)
self._outer_values_update(self.full_values)
def _raw_predict(self, Xnew, full_cov=False, kern=None):
"""