apunkt/GPy
1
0
Fork 0
mirror of https://github.com/SheffieldML/GPy.git synced 2026-05-12 05:22:38 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

Merge branch 'params' of github.com:SheffieldML/GPy into params

Browse source 
Conflicts:
	GPy/inference/latent_function_inference/var_dtc.py
This commit is contained in:
Ricardo 2014-03-17 19:05:37 +00:00
commit 9e64f116d8
3 changed files with 7 additions and 7 deletions
Download patch file Download diff file Expand all files Collapse all files

10
GPy/inference/latent_function_inference/var_dtc.py
View file

@ -65,9 +65,7 @@ class VarDTC(object):
_, output_dim = Y.shape _, output_dim = Y.shape
#see whether we've got a different noise variance for each datum #see whether we've got a different noise variance for each datum
#beta = 1./np.fmax(likelihood.variance, 1e-6)
beta = 1./np.fmax(likelihood.gaussian_variance(Y_metadata), 1e-6) beta = 1./np.fmax(likelihood.gaussian_variance(Y_metadata), 1e-6)
# VVT_factor is a matrix such that tdot(VVT_factor) = VVT...this is for efficiency! # VVT_factor is a matrix such that tdot(VVT_factor) = VVT...this is for efficiency!
#self.YYTfactor = self.get_YYTfactor(Y) #self.YYTfactor = self.get_YYTfactor(Y)
#VVT_factor = self.get_VVTfactor(self.YYTfactor, beta) #VVT_factor = self.get_VVTfactor(self.YYTfactor, beta)
@ -223,7 +221,7 @@ class VarDTCMissingData(object):
psi2_all = None psi2_all = None
Ys, traces = self._Y(Y) Ys, traces = self._Y(Y)
beta_all = 1./np.fmax(likelihood.variance, 1e-6) beta_all = 1./np.fmax(likelihood.gaussian_variance(Y_metadata), 1e-6)
het_noise = beta_all.size != 1 het_noise = beta_all.size != 1
import itertools import itertools
@ -330,18 +328,20 @@ class VarDTCMissingData(object):
diag.add(Bi, 1) diag.add(Bi, 1)
woodbury_inv_all[:, :, ind] = backsub_both_sides(Lm, Bi)[:,:,None] woodbury_inv_all[:, :, ind] = backsub_both_sides(Lm, Bi)[:,:,None]
dL_dthetaL = likelihood.exact_inference_gradients(dL_dR)
# gradients: # gradients:
if uncertain_inputs: if uncertain_inputs:
grad_dict = {'dL_dKmm': dL_dKmm, grad_dict = {'dL_dKmm': dL_dKmm,
'dL_dpsi0':dL_dpsi0_all, 'dL_dpsi0':dL_dpsi0_all,
'dL_dpsi1':dL_dpsi1_all, 'dL_dpsi1':dL_dpsi1_all,
'dL_dpsi2':dL_dpsi2_all, 'dL_dpsi2':dL_dpsi2_all,
'dL_dR':dL_dR} 'dL_dthetaL':dL_dthetaL}
else: else:
grad_dict = {'dL_dKmm': dL_dKmm, grad_dict = {'dL_dKmm': dL_dKmm,
'dL_dKdiag':dL_dpsi0_all, 'dL_dKdiag':dL_dpsi0_all,
'dL_dKnm':dL_dpsi1_all, 'dL_dKnm':dL_dpsi1_all,
'dL_dR':dL_dR} 'dL_dthetaL':dL_dthetaL}
#get sufficient things for posterior prediction #get sufficient things for posterior prediction
#TODO: do we really want to do this in the loop? #TODO: do we really want to do this in the loop?

2
GPy/models/bayesian_gplvm.py
View file

@ -66,7 +66,7 @@ class BayesianGPLVM(SparseGP):
super(BayesianGPLVM, self).parameters_changed() super(BayesianGPLVM, self).parameters_changed()
self._log_marginal_likelihood -= self.variational_prior.KL_divergence(self.X) self._log_marginal_likelihood -= self.variational_prior.KL_divergence(self.X)
self.X.mean.gradient, self.X.variance.gradient = self.kern.gradients_qX_expectations(variational_posterior=self.X, Z=self.Z, **self.grad_dict) self.X.mean.gradient, self.X.variance.gradient = self.kern.gradients_qX_expectations(variational_posterior=self.X, Z=self.Z, dL_dpsi0=self.grad_dict['dL_dpsi0'], dL_dpsi1=self.grad_dict['dL_dpsi1'], dL_dpsi2=self.grad_dict['dL_dpsi2'])
# update for the KL divergence # update for the KL divergence
self.variational_prior.update_gradients_KL(self.X) self.variational_prior.update_gradients_KL(self.X)

2
GPy/models/ss_gplvm.py
View file

@ -61,7 +61,7 @@ class SSGPLVM(SparseGP):
super(SSGPLVM, self).parameters_changed() super(SSGPLVM, self).parameters_changed()
self._log_marginal_likelihood -= self.variational_prior.KL_divergence(self.X) self._log_marginal_likelihood -= self.variational_prior.KL_divergence(self.X)
self.X.mean.gradient, self.X.variance.gradient, self.X.binary_prob.gradient = self.kern.gradients_qX_expectations(variational_posterior=self.X, Z=self.Z, **self.grad_dict) self.X.mean.gradient, self.X.variance.gradient, self.X.binary_prob.gradient = self.kern.gradients_qX_expectations(variational_posterior=self.X, Z=self.Z, dL_dpsi0=self.grad_dict['dL_dpsi0'], dL_dpsi1=self.grad_dict['dL_dpsi1'], dL_dpsi2=self.grad_dict['dL_dpsi2'])
# update for the KL divergence # update for the KL divergence
self.variational_prior.update_gradients_KL(self.X) self.variational_prior.update_gradients_KL(self.X)