add method for conditional log prob

dli_gpr.py

@@ -140,6 +140,13 @@ def conditional_distribution(self, new_y):
         #print(k.T @ precision @ k)
         return mean, cov
 
+    def conditional_log_prob(self, new_y, new_t):
+        """Compute log probability of out of sample points based on conditional distribution
+        """
+        mean, cov = self.conditional_distribution(new_y)
+        mvn = dist.MultivariateNormal(mean, covariance_matrix = cov + self.sigma * torch.eye(len(new_y)))
+        return mvn.log_prob(new_t)
+
 
 class dli_gpr:
     """ Implementation of Gaussian Process regression with non-isotropic noise
@@ -195,18 +202,11 @@ def initialize_variables(self, jitter=1e-5):
         self.scale_tril = torch.cholesky(self.scale + torch.eye(self.n) * jitter)
 
         # precision
-        self.beta = self.cluster_sizes * self.n / torch.sum(self.cluster_sizes)
-        self.beta_inverse = 1/self.beta
-
-        # compute inverse cluster sizes for gamma prior
-        #inverse_cluster_sizes = 1./self.cluster_sizes
-
-        # rate parameter for Gamma prior (beta inverse ~ variance of observations. Want lower beta_inverse for bigger clusters)
-        # initialize beta so that the MEAN of the gamma prior is proportional to the inverse weights of best linear unbiased estimator
-        #self.beta_inverse = inverse_cluster_sizes / torch.sum(inverse_cluster_sizes) * self.n
-
-        # beta is proportional to the expected precision 
-        #self.beta = 1./self.beta_inverse
+        inverse_cluster_sizes = 1./self.cluster_sizes
+        self.beta_inverse = inverse_cluster_sizes / torch.sum(inverse_cluster_sizes) * self.n
+        self.beta = 1./self.beta_inverse
+        #self.beta = self.cluster_sizes * self.n / torch.sum(self.cluster_sizes)
+        #self.beta_inverse = 1/self.beta
 
     def model(self):
         """Generative process"""
@@ -284,4 +284,11 @@ def conditional_distribution(self, new_y):
 
         return mean, cov
 
+    def conditional_log_prob(self, new_y, new_t):
+        """Compute log probability of out of sample points based on conditional distribution
+        """
+        mean, cov = self.conditional_distribution(new_y)
+        mvn = dist.MultivariateNormal(mean, covariance_matrix = cov)
+        return mvn.log_prob(new_t)
+