Added the structure to posterior.py to enable...

to enable computation from the posterior mean and variance, instead of
the woodbury componnents

This iss the first step in being able to use this structre for EP and
the laplace approximation.

GPy/core/gp_base.py

@@ -183,11 +183,10 @@ class GPBase(Model):
                 m, v = self._raw_predict(Xgrid, which_parts=which_parts)
                 lower = m - 2*np.sqrt(v)
                 upper = m + 2*np.sqrt(v)
-                Y = self.likelihood.Y
+                Y = self.Y
             else:
-                m, v, lower, upper = self.predict(Xgrid, which_parts=which_parts, sampling=False) #Compute the exact mean
+                m, v, lower, upper = self.predict(Xgrid, which_parts=which_parts) #Compute the exact mean
-                m_, v_, lower, upper = self.predict(Xgrid, which_parts=which_parts, sampling=True, num_samples=15000) #Apporximate the percentiles
+                Y = self.Y
-                Y = self.likelihood.data
             for d in which_data_ycols:
                 gpplot(Xnew, m[:, d], lower[:, d], upper[:, d], axes=ax, edgecol=linecol, fillcol=fillcol)
                 ax.plot(Xu[which_data_rows,free_dims], Y[which_data_rows, d], 'kx', mew=1.5)

GPy/inference/latent_function_inference/posterior.py

@@ -13,7 +13,7 @@ class Posterior(object):
     schemes and the model classes. 
 
     """
-    def __init__(self, log_marginal, dL_dK, dL_dtheta_lik, woodbury_chol, woodbury_vector, K):
+    def __init__(self, log_marginal, dL_dK, dL_dtheta_lik, woodbury_chol=None, woodbury_vector=None, K=None, mean=None, cov=None, K_chol=None):
         """
         log_marginal: log p(Y|X)
         DL_dK: d/dK log p(Y|X)
@@ -21,17 +21,51 @@ class Posterior(object):
         woodbury_chol : a lower triangular matrix L that satisfies posterior_covariance = K - K L^{-T} L^{-1} K
         woodbury_vector : a matrix (or vector, as Nx1 matrix) M which satisfies posterior_mean = K M
         K : the proir covariance (required for lazy computation of various quantities)
+        mean : the posterior mean
+        cov : the posterior covariance
+
+        Not all of the above need to be supplied! You *must* supply:
+          
+          log_marginal
+          dL_dK
+          dL_dtheta_lik
+          K (for lazy computation)
+
+       You may supply either:
+         
+          woodbury_chol
+          woodbury_vector
+
+        Or:
+          
+          mean
+          cov
+          K_chol (for lazy computation)
+
+        From the supplied quantities, all of the others will be computed on demand (lazy computation)
+          
         """
+        #obligatory
         self.log_marginal = log_marginal
         self.dL_dK = dL_dK
         self.dL_dtheta_lik = dL_dtheta_lik
-        self._woodbury_chol = woodbury_chol
-        self._woodbury_vector = woodbury_vector
         self._K = K
 
-        #these are computed lazily below
+        if ((woodbury_chol is not None) and (woodbury_vector is not None) and (K is not None)) or ((mean is not None) and (cov is not None) and (K is not None)):
-        self._mean = None
+            pass # we have sufficient to compute the posterior
-        self._covariance = None
+        else:
+            raise ValueError, "insufficient onformation to compute the posterior"
+
+        #option 1:
+        self._woodbury_chol = woodbury_chol
+        self._woodbury_vector = woodbury_vector
+
+        #option 2:
+        self._mean = mean
+        self._covariance = cov
+        self._K_chol = K_chol
+
+        #copmute this lazily
         self._precision = None
 
     @property
@@ -53,6 +87,20 @@ class Posterior(object):
             self._precision = np.linalg.inv(self.covariance)
         return self._precision
 
+    @property
+    def woodbury_chol(self):
+        if self._woodbury_chol is None:
+            ???
+        else:
+            return self._woodbury_chol
+
+    @property
+    def woodbury_vector(self):
+        if self._woodbury_vector is None:
+            ???
+        else:
+            return self._woodbury_vector
+
 
 
 

GPy/models/gp_regression.py

@@ -27,7 +27,7 @@ class GPRegression(GP):
 
         likelihood = likelihoods.Gaussian()
 
-        super(GPRegression, self).__init__(X, Y, kernel, likelihood, name='gp regression')
+        super(GPRegression, self).__init__(X, Y, kernel, likelihood, name='gp_regression')
 
     def getstate(self):
         return GP.getstate(self)