diff --git a/GPy/core/sparse_gp.py b/GPy/core/sparse_gp.py
index 56400d91..7f2d688d 100644
--- a/GPy/core/sparse_gp.py
+++ b/GPy/core/sparse_gp.py
@@ -112,6 +112,7 @@ class SparseGP(GP):
             #gradients wrt Z
             self.Z.gradient = self.kern.gradients_X(self.grad_dict['dL_dKmm'], self.Z)
             self.Z.gradient += self.kern.gradients_X(self.grad_dict['dL_dKnm'].T, self.Z, self.X)
+        self._Zgrad = self.Z.gradient.copy()
 
 
     def _raw_predict(self, Xnew, full_cov=False, kern=None):
diff --git a/GPy/examples/dimensionality_reduction.py b/GPy/examples/dimensionality_reduction.py
index 745c2c24..9d6686f9 100644
--- a/GPy/examples/dimensionality_reduction.py
+++ b/GPy/examples/dimensionality_reduction.py
@@ -405,10 +405,8 @@ def mrd_simulation(optimize=True, verbose=True, plot=True, plot_sim=True, **kw):
 
     D1, D2, D3, N, num_inducing, Q = 60, 20, 36, 60, 6, 5
     _, _, Ylist = _simulate_sincos(D1, D2, D3, N, num_inducing, plot_sim)
-    
 
-    # Ylist = [Ylist[0]]
-    k = kern.Linear(Q, ARD=True)
+    k = kern.Linear(Q) + kern.White(Q, variance=1e-4)
     m = MRD(Ylist, input_dim=Q, num_inducing=num_inducing, kernel=k, initx="PCA_concat", initz='permute', **kw)
 
     m['.*noise'] = [Y.var() / 40. for Y in Ylist]
@@ -428,8 +426,7 @@ def mrd_simulation_missing_data(optimize=True, verbose=True, plot=True, plot_sim
     D1, D2, D3, N, num_inducing, Q = 60, 20, 36, 60, 6, 5
     _, _, Ylist = _simulate_matern(D1, D2, D3, N, num_inducing, plot_sim)
 
-    # Ylist = [Ylist[0]]
-    k = kern.Linear(Q, ARD=True)
+    k = kern.Linear(Q) + kern.White(Q, variance=1e-4)
     inanlist = []
 
     for Y in Ylist:
diff --git a/GPy/models/bayesian_gplvm.py b/GPy/models/bayesian_gplvm.py
index 3ac703fe..88123227 100644
--- a/GPy/models/bayesian_gplvm.py
+++ b/GPy/models/bayesian_gplvm.py
@@ -97,7 +97,7 @@ class BayesianGPLVM(SparseGP_MPI):
                                             dL_dpsi2=self.grad_dict['dL_dpsi2'])
 
         self.variational_prior.update_gradients_KL(self.X)
-
+        self._Xgrad = self.X.gradient.copy()
 
         #super(BayesianGPLVM, self).parameters_changed()
         #self._log_marginal_likelihood -= self.variational_prior.KL_divergence(self.X)
diff --git a/GPy/models/mrd.py b/GPy/models/mrd.py
index cb98e1a8..8f098a1b 100644
--- a/GPy/models/mrd.py
+++ b/GPy/models/mrd.py
@@ -15,6 +15,7 @@ from ..util.initialization import initialize_latent
 from ..core.sparse_gp import SparseGP, GP
 from GPy.core.parameterization.variational import VariationalPosterior
 from GPy.models.bayesian_gplvm_minibatch import BayesianGPLVMMiniBatch
+from GPy.models.bayesian_gplvm import BayesianGPLVM
 from GPy.models.sparse_gp_minibatch import SparseGPMiniBatch
 
 class MRD(BayesianGPLVMMiniBatch):
@@ -170,7 +171,8 @@ class MRD(BayesianGPLVMMiniBatch):
             self._log_marginal_likelihood += b._log_marginal_likelihood
 
             self.logger.info('working on im <{}>'.format(hex(id(i))))
-            self.Z.gradient[:] += b.Z.gradient#full_values['Zgrad']
+            self.Z.gradient[:] += b._Zgrad  # b.Z.gradient  # full_values['Zgrad']
+
             #grad_dict = b.full_values
 
             if self.has_uncertain_inputs():
diff --git a/GPy/models/sparse_gp_minibatch.py b/GPy/models/sparse_gp_minibatch.py
index be068a0e..54160e6f 100644
--- a/GPy/models/sparse_gp_minibatch.py
+++ b/GPy/models/sparse_gp_minibatch.py
@@ -321,3 +321,4 @@ class SparseGPMiniBatch(SparseGP):
         else:
             self.posterior, self._log_marginal_likelihood, self.grad_dict = self._inner_parameters_changed(self.kern, self.X, self.Z, self.likelihood, self.Y_normalized, self.Y_metadata)
             self._outer_values_update(self.grad_dict)
+        self._Zgrad = self.Z.gradient.copy()