predictive_values implemented in EP

GPy/likelihoods/EP.py

@@ -18,7 +18,6 @@ class EP:
         self.likelihood_function = likelihood_function
         self.epsilon = epsilon
         self.eta, self.delta = power_ep
-        self.jitter = 1e-12 # TODO: is this needed?
 
         """
         Initial values - Likelihood approximation parameters:
@@ -27,6 +26,16 @@ class EP:
         self.tau_tilde = np.zeros(self.N)
         self.v_tilde = np.zeros(self.N)
 
+    def predictive_values(self,mu,var):
+        return self.likelihood_function.predictive_values(mu,var)
+
+    def _get_params(self):
+        return np.zeros(0)
+    def _get_param_names(self):
+        return []
+    def _set_params(self,p):
+        pass # TODO: the EP likelihood might want to take some parameters...
+
     def _compute_GP_variables(self):
         #Variables to be called from GP
         mu_tilde = self.v_tilde/self.tau_tilde #When calling EP, this variable is used instead of Y in the GP model

GPy/likelihoods/Gaussian.py

@@ -29,6 +29,16 @@ class Gaussian:
         self._variance = x
         self.variance = np.eye(self.N)*self._variance
 
+    def predictive_values(self,mu,var):
+        """
+        Un-normalise the prediction and add the likelihood variance, then return the 5%, 95% interval
+        """
+        mean = mu*self._std + self._mean
+        true_var = (var + self._variance)*self._std**2
+        _5pc = mean + mean - 2.*np.sqrt(var)
+        _95pc = mean + 2.*np.sqrt(var)
+        return mean, _5pc, _95pc
+
     def fit(self):
         """
         No approximations needed

GPy/likelihoods/likelihood_functions.py

@@ -49,7 +49,7 @@ class probit(likelihood):
 
     def predictive_values(self,mu,var,all=False):
         """
-        Compute  mean, variance, and conficence interval (percentiles 5 and 95) of the  prediction
+        Compute  mean, and conficence interval (percentiles 5 and 95) of the  prediction
         """
         mu = mu.flatten()
         var = var.flatten()
@@ -57,7 +57,7 @@ class probit(likelihood):
         if all:
             p_05 = np.zeros([mu.size])
             p_95 = np.ones([mu.size])
-            return mean, mean*(1-mean),p_05,p_95
+            return mean, p_05, p_95
         else:
             return mean
 
@@ -136,14 +136,14 @@ class poisson(likelihood):
 
     def predictive_values(self,mu,var,all=False):
         """
-        Compute  mean, variance, and conficence interval (percentiles 5 and 95) of the  prediction
+        Compute  mean, and conficence interval (percentiles 5 and 95) of the  prediction
         """
         mean = np.exp(mu*self.scale + self.location)
         if all:
             tmp = stats.poisson.ppf(np.array([.05,.95]),mu)
             p_05 = tmp[:,0]
             p_95 = tmp[:,1]
-            return mean,mean,p_05,p_95
+            return mean,p_05,p_95
         else:
             return mean
 

GPy/models/GP.py

@@ -164,9 +164,10 @@ class GP(model):
         """
         #normalise X values
         Xnew = (Xnew.copy() - self._Xmean) / self._Xstd
-        mu, var, phi = self._raw_predict(Xnew, slices, full_cov=full_cov)
+        mu, var = self._raw_predict(Xnew, slices, full_cov=full_cov)
 
         #now push through likelihood TODO
+        mean, _5pc, _95pc = self.likelihood.predictive_values(mu, var)
 
         return mean, _5pc, _95pc