minor fixes

candlegp/kernels.py

@@ -198,15 +198,15 @@ def __init__(self, input_dim, variance=1.0, active_dims=None, name=None):
         super(Static, self).__init__(input_dim, active_dims, name=name)
         self.variance = parameter.PositiveParam(variance)
 
-    def Kdiag(self, X):
+    def Kdiag(self, X, presliced=False):
         return self.variance.get().expand(X.size(0))
 
 
 class White(Static):
     """
     The White kernel
     """
-    #@params_as_tensors
+
     def K(self, X, X2=None, presliced=False):
         if X2 is None:
             d = self.variance.get().expand(X.size(0))

candlegp/models/__init__.py

@@ -4,3 +4,4 @@
 from .svgp import SVGP
 from .vgp import VGP
 from .gpmc import GPMC
+from .sgpmc import SGPMC

candlegp/models/gpmc.py

@@ -52,7 +52,7 @@ def __init__(self, X, Y, kern, likelihood,
         self.num_data = X.size(0)
         self.num_latent = num_latent or Y.size(1)
         self.V = parameter.Param(self.X.data.new(self.num_data, self.num_latent).zero_())
-        self.V.prior = priors.Gaussian(0., 1.)
+        self.V.prior = priors.Gaussian(self.X.data.new(1).fill_(0.), self.X.data.new(1).fill_(1.))
 
     def compute_log_likelihood(self):
         """
@@ -80,7 +80,7 @@ def predict_f(self, Xnew, full_cov=False):
         where F* are points on the GP at Xnew, F=LV are points on the GP at X.
 
         """
-        mu, var = conditionals.conditional(Xnew, self.X, self.kern, self.V,
+        mu, var = conditionals.conditional(Xnew, self.X, self.kern, self.V.get(),
                               full_cov=full_cov,
                               q_sqrt=None, whiten=True)
         return mu + self.mean_function(Xnew), var

candlegp/models/sgpr.py

@@ -146,7 +146,7 @@ def compute_log_likelihood(self):
         # compute log marginal bound
         bound = -0.5 * num_data * output_dim * float(numpy.log(2 * numpy.pi))
         bound += -output_dim * torch.sum(torch.log(torch.diag(LB)))
-        bound -= 0.5 * num_data * output_dim * torch.log(self.likelihood.variance.get())
+        bound = bound - 0.5 * num_data * output_dim * torch.log(self.likelihood.variance.get())
         bound += -0.5 * torch.sum(err**2) / self.likelihood.variance.get()
         bound += 0.5 * torch.sum(c**2)
         bound += -0.5 * output_dim * torch.sum(Kdiag) / self.likelihood.variance.get()

candlegp/priors.py

@@ -22,7 +22,7 @@
 def wrap(x, ttype=torch.Tensor, **argd):
     if numpy.isscalar(x):
         x = Variable(ttype([x]),**argd)
-    elif isinstance(x, [torch.Tensor, torch.DoubleTensor]):
+    elif isinstance(x, (torch.Tensor, torch.DoubleTensor)):
         x = Variable(x, **argd)
     return x