apunkt/GPy
1
0
Fork 0
mirror of https://github.com/SheffieldML/GPy.git synced 2026-05-13 14:03:20 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

[whitespaces] & [opt] minor fix of optimizer, when Optimizer is provided (set model to self)

Browse source
This commit is contained in:
mzwiessele 2014-06-20 10:27:52 -07:00
parent eaa18e3512
commit 73d9118043
2 changed files with 11 additions and 11 deletions
Download patch file Download diff file Expand all files Collapse all files

1
GPy/core/model.py
View file

@ -229,6 +229,7 @@ class Model(Parameterized):
optimizer = self.preferred_optimizer optimizer = self.preferred_optimizer
elif isinstance(optimizer, optimization.Optimizer): elif isinstance(optimizer, optimization.Optimizer):
opt = optimizer opt = optimizer
opt.model = self
else: else:
optimizer = optimization.get_optimizer(optimizer) optimizer = optimization.get_optimizer(optimizer)
opt = optimizer(start, model=self, **kwargs) opt = optimizer(start, model=self, **kwargs)

21
GPy/models/bayesian_gplvm.py
View file

@ -49,7 +49,6 @@ class BayesianGPLVM(SparseGP):
if likelihood is None: if likelihood is None:
likelihood = Gaussian() likelihood = Gaussian()
self.variational_prior = NormalPrior() self.variational_prior = NormalPrior()
X = NormalPosterior(X, X_variance) X = NormalPosterior(X, X_variance)
@ -75,7 +74,7 @@ class BayesianGPLVM(SparseGP):
if isinstance(self.inference_method, VarDTC_GPU): if isinstance(self.inference_method, VarDTC_GPU):
update_gradients(self) update_gradients(self)
return return
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)
@ -87,7 +86,7 @@ class BayesianGPLVM(SparseGP):
def plot_latent(self, labels=None, which_indices=None, def plot_latent(self, labels=None, which_indices=None,
resolution=50, ax=None, marker='o', s=40, resolution=50, ax=None, marker='o', s=40,
fignum=None, plot_inducing=True, legend=True, fignum=None, plot_inducing=True, legend=True,
plot_limits=None, plot_limits=None,
aspect='auto', updates=False, predict_kwargs={}, imshow_kwargs={}): aspect='auto', updates=False, predict_kwargs={}, imshow_kwargs={}):
import sys import sys
assert "matplotlib" in sys.modules, "matplotlib package has not been imported." assert "matplotlib" in sys.modules, "matplotlib package has not been imported."
@ -107,10 +106,10 @@ class BayesianGPLVM(SparseGP):
""" """
N_test = Y.shape[0] N_test = Y.shape[0]
input_dim = self.Z.shape[1] input_dim = self.Z.shape[1]
means = np.zeros((N_test, input_dim)) means = np.zeros((N_test, input_dim))
covars = np.zeros((N_test, input_dim)) covars = np.zeros((N_test, input_dim))
dpsi0 = -0.5 * self.input_dim / self.likelihood.variance dpsi0 = -0.5 * self.input_dim / self.likelihood.variance
dpsi2 = self.grad_dict['dL_dpsi2'][0][None, :, :] # TODO: this may change if we ignore het. likelihoods dpsi2 = self.grad_dict['dL_dpsi2'][0][None, :, :] # TODO: this may change if we ignore het. likelihoods
V = Y/self.likelihood.variance V = Y/self.likelihood.variance
@ -125,7 +124,7 @@ class BayesianGPLVM(SparseGP):
dpsi1 = np.dot(self.posterior.woodbury_vector, V.T) dpsi1 = np.dot(self.posterior.woodbury_vector, V.T)
#start = np.zeros(self.input_dim * 2) #start = np.zeros(self.input_dim * 2)
from scipy.optimize import minimize from scipy.optimize import minimize
@ -139,7 +138,7 @@ class BayesianGPLVM(SparseGP):
X = NormalPosterior(means, covars) X = NormalPosterior(means, covars)
return X return X
def dmu_dX(self, Xnew): def dmu_dX(self, Xnew):
""" """
@ -169,7 +168,7 @@ class BayesianGPLVM(SparseGP):
from ..plotting.matplot_dep import dim_reduction_plots from ..plotting.matplot_dep import dim_reduction_plots
return dim_reduction_plots.plot_steepest_gradient_map(self,*args,**kwargs) return dim_reduction_plots.plot_steepest_gradient_map(self,*args,**kwargs)
def latent_cost_and_grad(mu_S, input_dim, kern, Z, dL_dpsi0, dL_dpsi1, dL_dpsi2): def latent_cost_and_grad(mu_S, input_dim, kern, Z, dL_dpsi0, dL_dpsi1, dL_dpsi2):
""" """
@ -187,10 +186,10 @@ def latent_cost_and_grad(mu_S, input_dim, kern, Z, dL_dpsi0, dL_dpsi1, dL_dpsi2)
psi2 = kern.psi2(Z, X) psi2 = kern.psi2(Z, X)
lik = dL_dpsi0 * psi0.sum() + np.einsum('ij,kj->...', dL_dpsi1, psi1) + np.einsum('ijk,lkj->...', dL_dpsi2, psi2) - 0.5 * np.sum(np.square(mu) + S) + 0.5 * np.sum(log_S) lik = dL_dpsi0 * psi0.sum() + np.einsum('ij,kj->...', dL_dpsi1, psi1) + np.einsum('ijk,lkj->...', dL_dpsi2, psi2) - 0.5 * np.sum(np.square(mu) + S) + 0.5 * np.sum(log_S)
dLdmu, dLdS = kern.gradients_qX_expectations(dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, X) dLdmu, dLdS = kern.gradients_qX_expectations(dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, X)
dmu = dLdmu - mu dmu = dLdmu - mu
# dS = S0 + S1 + S2 -0.5 + .5/S # dS = S0 + S1 + S2 -0.5 + .5/S
dlnS = S * (dLdS - 0.5) + .5 dlnS = S * (dLdS - 0.5) + .5
return -lik, -np.hstack((dmu.flatten(), dlnS.flatten())) return -lik, -np.hstack((dmu.flatten(), dlnS.flatten()))