consider MRD layers in prediction method

Author: vcuculo

deepgp/layers/mrd.py

@@ -21,7 +21,7 @@ def __init__(self, Y, dim_down, dim_up, likelihood, MLP_dims=None, X=None, X_var
         self.scale = 1.
 
         if back_constraint:
-            from .mlp import MLP
+            from deepgp.encoder.mlp import MLP
             from copy import deepcopy
             self.encoder = MLP([dim_down, int((dim_down+dim_up)*2./3.), int((dim_down+dim_up)/3.), dim_up] if MLP_dims is None else [dim_down]+deepcopy(MLP_dims)+[dim_up])
             X = self.encoder.predict(Y.mean.values if isinstance(Y, VariationalPosterior) else Y)

deepgp/models/model.py

@@ -338,7 +338,7 @@ def collect_all_XY(self, root=0):
             if self.mpi_comm.rank==root: return XY
             else: return None
 
-    def predict(self, Xnew, full_cov=False, Y_metadata=None, kern=None):
+    def predict(self, Xnew, full_cov=False, Y_metadata=None, kern=None, view=0):
         """Make a prediction from the deep Gaussian process model for a given input"""
         from GPy.core.parameterization.variational import NormalPosterior
 
@@ -359,7 +359,11 @@ def predict(self, Xnew, full_cov=False, Y_metadata=None, kern=None):
                 if var.shape[1]==1:
                     var = np.tile(var,mean.shape[1])
                 x = NormalPosterior(mean, var)
-        return self.layers[0].predict(x)
+        mrd_flag = hasattr(self.layers[0], 'views')
+        if mrd_flag:
+            return self.layers[0].views[view].predict(x)
+        else:
+            return self.layers[0].predict(x)
 
     @Model.optimizer_array.setter
     def optimizer_array(self, p):

examples/example_supervised_learning.py

@@ -46,7 +46,7 @@
 # Number of inducing points to use
 num_inducing = 40
 # Whether to use back-constraint for variational posterior
-back_constraint = False
+back_constraint = True
 # Dimensions of the MLP back-constraint if set to true
 encoder_dims=[[300],[150]]
 

examples/example_unsupervised_learning.py

@@ -37,7 +37,7 @@
     output_var = m.layers[i].Y.var() if i==0 else m.layers[i].Y.mean.var()
     m.layers[i].Gaussian_noise.variance = output_var*0.01
 
-m.optimize(max_iters=5000, messages=True)
+m.optimize(max_iters=500, messages=True)
 
 
 #--------- Inspection ----------#
@@ -50,4 +50,5 @@
 plt.figure(figsize=(5,5))
 deepgp.util.visualize_DGP(m, labels, layer=0, dims=[1,2]); plt.title('Layer 0')
 plt.figure(figsize=(5,5))
-deepgp.util.visualize_DGP(m, labels, layer=1, dims=[0,1]); plt.title('Layer 1')
+deepgp.util.visualize_DGP(m, labels, layer=1, dims=[0,1]); plt.title('Layer 1')
+plt.show()

files.txt

@@ -0,0 +1 @@
+/usr/local/lib/python2.7/dist-packages/DGP-1.0-py2.7.egg

setup.py

@@ -20,6 +20,7 @@ def read(fname):
       keywords = "deep learning",
       url = "",
       packages = ["deepgp",
+                  "deepgp.encoder",
                   "deepgp.inference",
                   "deepgp.layers",
                   "deepgp.util",