apunkt/GPy
1
0
Fork 0
mirror of https://github.com/SheffieldML/GPy.git synced 2026-05-27 14:25:16 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

variational returns now the right raveled indices

Browse source
This commit is contained in:
Max Zwiessele 2014-03-24 09:06:48 +00:00
parent ae81cbce6c
commit ed2aaab4bb
1 changed files with 15 additions and 9 deletions
Download patch file Download diff file Expand all files Collapse all files

24
GPy/core/parameterization/variational.py
View file

@ -12,7 +12,7 @@ from transformations import Logexp, Logistic
class VariationalPrior(Parameterized): class VariationalPrior(Parameterized):
def __init__(self, name='latent space', **kw): def __init__(self, name='latent space', **kw):
super(VariationalPrior, self).__init__(name=name, **kw) super(VariationalPrior, self).__init__(name=name, **kw)
def KL_divergence(self, variational_posterior): def KL_divergence(self, variational_posterior):
raise NotImplementedError, "override this for variational inference of latent space" raise NotImplementedError, "override this for variational inference of latent space"
@ -21,8 +21,8 @@ class VariationalPrior(Parameterized):
updates the gradients for mean and variance **in place** updates the gradients for mean and variance **in place**
""" """
raise NotImplementedError, "override this for variational inference of latent space" raise NotImplementedError, "override this for variational inference of latent space"
class NormalPrior(VariationalPrior): class NormalPrior(VariationalPrior):
def KL_divergence(self, variational_posterior): def KL_divergence(self, variational_posterior):
var_mean = np.square(variational_posterior.mean).sum() var_mean = np.square(variational_posterior.mean).sum()
var_S = (variational_posterior.variance - np.log(variational_posterior.variance)).sum() var_S = (variational_posterior.variance - np.log(variational_posterior.variance)).sum()
@ -40,7 +40,7 @@ class SpikeAndSlabPrior(VariationalPrior):
self.pi = Param('pi', pi, Logistic(1e-10,1.-1e-10)) self.pi = Param('pi', pi, Logistic(1e-10,1.-1e-10))
self.variance = Param('variance',variance) self.variance = Param('variance',variance)
self.add_parameters(self.pi) self.add_parameters(self.pi)
def KL_divergence(self, variational_posterior): def KL_divergence(self, variational_posterior):
mu = variational_posterior.mean mu = variational_posterior.mean
S = variational_posterior.variance S = variational_posterior.variance
@ -49,7 +49,7 @@ class SpikeAndSlabPrior(VariationalPrior):
var_S = (S - np.log(S)) var_S = (S - np.log(S))
var_gamma = (gamma*np.log(gamma/self.pi)).sum()+((1-gamma)*np.log((1-gamma)/(1-self.pi))).sum() var_gamma = (gamma*np.log(gamma/self.pi)).sum()+((1-gamma)*np.log((1-gamma)/(1-self.pi))).sum()
return var_gamma+ 0.5 * (gamma* (var_mean + var_S -1)).sum() return var_gamma+ 0.5 * (gamma* (var_mean + var_S -1)).sum()
def update_gradients_KL(self, variational_posterior): def update_gradients_KL(self, variational_posterior):
mu = variational_posterior.mean mu = variational_posterior.mean
S = variational_posterior.variance S = variational_posterior.variance
@ -59,8 +59,6 @@ class SpikeAndSlabPrior(VariationalPrior):
mu.gradient -= gamma*mu mu.gradient -= gamma*mu
S.gradient -= (1. - (1. / (S))) * gamma /2. S.gradient -= (1. - (1. / (S))) * gamma /2.
self.pi.gradient = (gamma/self.pi - (1.-gamma)/(1.-self.pi)).sum(axis=0) self.pi.gradient = (gamma/self.pi - (1.-gamma)/(1.-self.pi)).sum(axis=0)
class VariationalPosterior(Parameterized): class VariationalPosterior(Parameterized):
def __init__(self, means=None, variances=None, name=None, *a, **kw): def __init__(self, means=None, variances=None, name=None, *a, **kw):
@ -74,7 +72,15 @@ class VariationalPosterior(Parameterized):
self.num_data, self.input_dim = self.mean.shape self.num_data, self.input_dim = self.mean.shape
if self.has_uncertain_inputs(): if self.has_uncertain_inputs():
assert self.variance.shape == self.mean.shape, "need one variance per sample and dimenion" assert self.variance.shape == self.mean.shape, "need one variance per sample and dimenion"
def _raveled_index(self):
index = np.empty(dtype=int, shape=0)
size = 0
for p in self._parameters_:
index = np.hstack((index, p._raveled_index()+size))
size += p._realsize_ if hasattr(p, '_realsize_') else p.size
return index
def has_uncertain_inputs(self): def has_uncertain_inputs(self):
return not self.variance is None return not self.variance is None
@ -126,7 +132,7 @@ class SpikeAndSlabPosterior(VariationalPosterior):
super(SpikeAndSlabPosterior, self).__init__(means, variances, name) super(SpikeAndSlabPosterior, self).__init__(means, variances, name)
self.gamma = Param("binary_prob",binary_prob, Logistic(1e-10,1.-1e-10)) self.gamma = Param("binary_prob",binary_prob, Logistic(1e-10,1.-1e-10))
self.add_parameter(self.gamma) self.add_parameter(self.gamma)
def __getitem__(self, s): def __getitem__(self, s):
if isinstance(s, (int, slice, tuple, list, np.ndarray)): if isinstance(s, (int, slice, tuple, list, np.ndarray)):
import copy import copy