some work on ep, and some messing with wher ethe derivatives are computed (in the model, not the inference

GPy/core/gp.py

@@ -70,7 +70,7 @@ class GP(Model):
 
     def parameters_changed(self):
         self.posterior, self._log_marginal_likelihood, grad_dict = self.inference_method.inference(self.kern, self.X, self.likelihood, self.Y, Y_metadata=self.Y_metadata)
-        self._dL_dK = grad_dict['dL_dK']
+        self.kern.update_gradients_full(grad_dict['dL_dK'])
 
     def log_likelihood(self):
         return self._log_marginal_likelihood

GPy/inference/latent_function_inference/ep.py

@@ -23,11 +23,26 @@ class EP(object):
         self.old_mutilde, self.old_vtilde = None, None
 
     def inference(self, kern, X, likelihood, Y, Y_metadata=None):
+        num_data, output_dim = X.shape
+        assert output_dim ==1, "ep in 1D only (for now!)"
 
         K = kern.K(X)
 
         mu_tilde, tau_tilde = self.expectation_propagation()
 
+        Wi, LW, LWi, W_logdet = pdinv(K + np.diag(1./tau_tilde)
+
+        alpha, _ = dpotrs(LW, mu_tilde, lower=1)
+
+        log_marginal =  0.5*(-num_data * log_2_pi - W_logdet - np.sum(alpha * mu_tilde))
+
+        dL_dK = 0.5 * (tdot(alpha[:,None]) - Wi)
+
+        #TODO: what abot derivatives of the likelihood parameters?
+
+        return Posterior(woodbury_inv=Wi, woodbury_vector=alpha, K=K), log_marginal, {'dL_dK':dL_dK}
+
+
 
     def expectation_propagation(self, K, Y, Y_metadata, likelihood)
 

GPy/inference/latent_function_inference/exact_gaussian_inference.py

@@ -49,8 +49,7 @@ class ExactGaussianInference(object):
 
         dL_dK = 0.5 * (tdot(alpha) - Y.shape[1] * Wi)
 
-        kern.update_gradients_full(dL_dK, X)
-
+        #TODO: does this really live here?
         likelihood.update_gradients(np.diag(dL_dK))
 
         return Posterior(woodbury_chol=LW, woodbury_vector=alpha, K=K), log_marginal, {'dL_dK':dL_dK}

GPy/util/multioutput.py

@@ -2,6 +2,27 @@ import numpy as np
 import warnings
 from .. import kern
 
+def build_XY(input_list,output_list=None,index=None):
+    num_outputs = len(input_list)
+    _s = [0] + [ _x.shape[0] for _x in input_list ]
+    _s = np.cumsum(_s)
+    slices = [slice(a,b) for a,b in zip(_s[:-1],_s[1:])]
+    if output_list is not None:
+        assert num_outputs == len(output_list)
+        Y = np.vstack(output_list)
+    else:
+        Y = None
+
+    if index is not None:
+        assert len(index) == num_outputs
+        I = np.vstack( [j*np.ones((_x.shape[0],1)) for _x,j in zip(input_list,index)] )
+    else:
+        I = np.vstack( [j*np.ones((_x.shape[0],1)) for _x,j in zip(input_list,range(num_outputs))] )
+
+    X = np.vstack(input_list)
+    X = np.hstack([X,I])
+    return X,Y,slices
+
 def build_lcm(input_dim, num_outputs, CK = [], NC = [], W_columns=1,W=None,kappa=None):
     #TODO build_icm or build_lcm
     """
@@ -25,9 +46,9 @@ def build_lcm(input_dim, num_outputs, CK = [], NC = [], W_columns=1,W=None,kappa
             k.input_dim = input_dim + 1
             warnings.warn("kernel's input dimension overwritten to fit input_dim parameter.")
 
-    kernel = CK[0].prod(kern.coregionalize(num_outputs,W_columns,W,kappa),tensor=True)
+    kernel = CK[0].prod(kern.Coregionalize(num_outputs,W_columns,W,kappa),tensor=True)
     for k in CK[1:]:
-        k_coreg = kern.coregionalize(num_outputs,W_columns,W,kappa)
+        k_coreg = kern.Coregionalize(num_outputs,W_columns,W,kappa)
         kernel += k.prod(k_coreg,tensor=True)
     for k in NC:
         kernel += k