apunkt/GPy
1
0
Fork 0
mirror of https://github.com/SheffieldML/GPy.git synced 2026-05-30 14:35:15 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

proper implementation of group_spike

Browse source
This commit is contained in:
Zhenwen Dai 2014-09-23 13:26:12 +01:00
parent 494f38f788
commit 4018832be5
2 changed files with 24 additions and 7 deletions
Download patch file Download diff file Expand all files Collapse all files

29
GPy/core/parameterization/variational.py
View file

@ -34,10 +34,11 @@ class NormalPrior(VariationalPrior):
variational_posterior.variance.gradient -= (1. - (1. / (variational_posterior.variance))) * 0.5 variational_posterior.variance.gradient -= (1. - (1. / (variational_posterior.variance))) * 0.5
class SpikeAndSlabPrior(VariationalPrior): class SpikeAndSlabPrior(VariationalPrior):
def __init__(self, pi=None, learnPi=False, variance = 1.0, name='SpikeAndSlabPrior', **kw): def __init__(self, pi=None, learnPi=False, group_spike=False, variance = 1.0, name='SpikeAndSlabPrior', **kw):
super(SpikeAndSlabPrior, self).__init__(name=name, **kw) super(SpikeAndSlabPrior, self).__init__(name=name, **kw)
self.variance = Param('variance',variance) self.variance = Param('variance',variance)
self.learnPi = learnPi self.learnPi = learnPi
self.group_spike = group_spike
if learnPi: if learnPi:
self.pi = Param('Pi', pi, Logistic(1e-10,1.-1e-10)) self.pi = Param('Pi', pi, Logistic(1e-10,1.-1e-10))
else: else:
@ -57,7 +58,11 @@ class SpikeAndSlabPrior(VariationalPrior):
var_mean = np.square(mu)/self.variance var_mean = np.square(mu)/self.variance
var_S = (S/self.variance - np.log(S)) var_S = (S/self.variance - np.log(S))
var_gamma = (gamma*np.log(gamma/pi)).sum()+((1-gamma)*np.log((1-gamma)/(1-pi))).sum() # TODO: sovle group_spike for parallelization
if self.group_spike:
var_gamma = (gamma*np.log(gamma/pi)).sum()/gamma.shape[0]+((1-gamma)*np.log((1-gamma)/(1-pi))).sum()/gamma.shape[0]
else:
var_gamma = (gamma*np.log(gamma/pi)).sum()+((1-gamma)*np.log((1-gamma)/(1-pi))).sum()
return var_gamma+ (gamma* (np.log(self.variance)-1. +var_mean + var_S)).sum()/2. return var_gamma+ (gamma* (np.log(self.variance)-1. +var_mean + var_S)).sum()/2.
def update_gradients_KL(self, variational_posterior): def update_gradients_KL(self, variational_posterior):
@ -70,16 +75,28 @@ class SpikeAndSlabPrior(VariationalPrior):
else: else:
pi = self.pi pi = self.pi
gamma.gradient -= np.log((1-pi)/pi*gamma/(1.-gamma))+((np.square(mu)+S)/self.variance-np.log(S)+np.log(self.variance)-1.)/2. if self.group_spike:
gamma.gradient -= np.log((1-pi)/pi*gamma/(1.-gamma))/gamma.shape[0]+((np.square(mu)+S)/self.variance-np.log(S)+np.log(self.variance)-1.)/2.
else:
gamma.gradient -= np.log((1-pi)/pi*gamma/(1.-gamma))+((np.square(mu)+S)/self.variance-np.log(S)+np.log(self.variance)-1.)/2.
mu.gradient -= gamma*mu/self.variance mu.gradient -= gamma*mu/self.variance
S.gradient -= (1./self.variance - 1./S) * gamma /2. S.gradient -= (1./self.variance - 1./S) * gamma /2.
if self.learnPi: if self.learnPi:
if len(self.pi)==1: if len(self.pi)==1:
self.pi.gradient = (gamma/self.pi - (1.-gamma)/(1.-self.pi)).sum() if self.group_spike:
self.pi.gradient = (gamma/self.pi - (1.-gamma)/(1.-self.pi)).sum()/gamma.shape[0]
else:
self.pi.gradient = (gamma/self.pi - (1.-gamma)/(1.-self.pi)).sum()
elif len(self.pi.shape)==1: elif len(self.pi.shape)==1:
self.pi.gradient = (gamma/self.pi - (1.-gamma)/(1.-self.pi)).sum(axis=0) if self.group_spike:
self.pi.gradient = (gamma/self.pi - (1.-gamma)/(1.-self.pi)).sum(axis=0)/gamma.shape[0]
else:
self.pi.gradient = (gamma/self.pi - (1.-gamma)/(1.-self.pi)).sum(axis=0)
else: else:
self.pi[idx].gradient = (gamma/self.pi[idx] - (1.-gamma)/(1.-self.pi[idx])) if self.group_spike:
self.pi[idx].gradient = (gamma/self.pi[idx] - (1.-gamma)/(1.-self.pi[idx]))/gamma.shape[0]
else:
self.pi[idx].gradient = (gamma/self.pi[idx] - (1.-gamma)/(1.-self.pi[idx]))
class VariationalPosterior(Parameterized): class VariationalPosterior(Parameterized):
def __init__(self, means=None, variances=None, name='latent space', *a, **kw): def __init__(self, means=None, variances=None, name='latent space', *a, **kw):

2
GPy/models/ss_gplvm.py
View file

@ -65,7 +65,7 @@ class SSGPLVM(SparseGP_MPI):
if pi is None: if pi is None:
pi = np.empty((input_dim)) pi = np.empty((input_dim))
pi[:] = 0.5 pi[:] = 0.5
self.variational_prior = SpikeAndSlabPrior(pi=pi,learnPi=learnPi) # the prior probability of the latent binary variable b self.variational_prior = SpikeAndSlabPrior(pi=pi,learnPi=learnPi, group_spike=group_spike) # the prior probability of the latent binary variable b
X = SpikeAndSlabPosterior(X, X_variance, gamma) X = SpikeAndSlabPosterior(X, X_variance, gamma)