apunkt/GPy
1
0
Fork 0
mirror of https://github.com/SheffieldML/GPy.git synced 2026-05-18 13:55:14 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

confidence interval fixed

Browse source
This commit is contained in:
Ricardo 2013-07-08 18:33:44 +01:00
parent b856c60d30
commit b3fd843d33
2 changed files with 18 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

16
GPy/likelihoods/noise_models/likelihood_functions.py
View file

@ -279,6 +279,18 @@ class NoiseModel(object):
var_exp = exp_exp2 - predictive_mean**2 var_exp = exp_exp2 - predictive_mean**2
return exp_var + var_exp return exp_var + var_exp
def _predictive_percentiles(self,p,mu,sigma):
"""
Percentiles of the predictive distribution
:parm p: lower tail probability
:param mu: cavity distribution mean
:param sigma: cavity distribution standard deviation
:predictive_mean: output's predictive mean, if None _predictive_mean function will be called.
"""
qf = stats.norm.ppf(p,mu,sigma)
return self.link.inv_transf(qf)
def _nlog_joint_predictive_scaled(self,x,mu,sigma): def _nlog_joint_predictive_scaled(self,x,mu,sigma):
""" """
Negative logarithm of the joint predictive distribution (latent variable and output). Negative logarithm of the joint predictive distribution (latent variable and output).
@ -340,8 +352,8 @@ class NoiseModel(object):
for m,s in zip(mu,np.sqrt(var)): for m,s in zip(mu,np.sqrt(var)):
pred_mean.append(self.predictive_mean(m,s)) pred_mean.append(self.predictive_mean(m,s))
pred_var.append(self.predictive_variance(m,s,pred_mean[-1])) pred_var.append(self.predictive_variance(m,s,pred_mean[-1]))
q1.append(self.predictive_mean(stats.norm.ppf(.025,m,s**2),s)) q1.append(self._predictive_percentiles(.025,m,s))
q3.append(self.predictive_mean(stats.norm.ppf(.975,m,s**2),s)) q3.append(self._predictive_percentiles(.975,m,s))
pred_mean = np.vstack(pred_mean) pred_mean = np.vstack(pred_mean)
pred_var = np.vstack(pred_var) pred_var = np.vstack(pred_var)
q1 = np.vstack(q1) q1 = np.vstack(q1)

8
GPy/likelihoods/noise_models/poisson_likelihood.py
View file

@ -27,15 +27,15 @@ class Poisson(NoiseModel):
super(Poisson, self).__init__(link,analytical_moments) super(Poisson, self).__init__(link,analytical_moments)
def _preprocess_values(self,Y): #TODO def _preprocess_values(self,Y): #TODO
self.scale = .5*Y.max() #self.scale = .5*Y.max()
self.shift = Y.mean() #self.shift = Y.mean()
return (Y - self.shift)/self.scale return Y #(Y - self.shift)/self.scale
def _mass(self,gp,obs): def _mass(self,gp,obs):
""" """
Mass (or density) function Mass (or density) function
""" """
obs = obs*self.scale + self.shift #obs = obs*self.scale + self.shift
return stats.poisson.pmf(obs,self.link.inv_transf(gp)) return stats.poisson.pmf(obs,self.link.inv_transf(gp))
def _nlog_mass(self,gp,obs): def _nlog_mass(self,gp,obs):