Hotfix 0.8.3

Merge branch 'hotfix-appnp'

CHANGELOG.md

@@ -1,5 +1,11 @@
 # Change Log
 
+## [0.8.3](https://github.com/stellargraph/stellargraph/tree/v0.8.3)
+
+**Fixed bugs:**
+- Fixed the issue in the APPNP class that causes appnp to propagate excessive dropout layers. [\#525](https://github.com/stellargraph/stellargraph/pull/525)
+- Added a fix into the PPNP node classification demo so that the softmax layer is no longer propagated.
+
 ## [0.8.2](https://github.com/stellargraph/stellargraph/tree/v0.8.2)
 
 **Fixed bugs:**

stellargraph/layer/appnp.py

@@ -180,7 +180,7 @@ class APPNP:
         bias (bool): toggles an optional bias in fully connected layers
         dropout (float): dropout rate applied to input features of each layer
         kernel_regularizer (str): normalization applied to the kernels of fully connetcted layers
-        teleport_probability: "probability" of returning to the starting node in the propogation step as desribed  in
+        teleport_probability: "probability" of returning to the starting node in the propagation step as desribed in
         the paper (alpha in the paper)
         approx_iter: number of iterations to approximate PPNP as described in the paper (K in the paper)
     """
@@ -305,12 +305,15 @@ def __call__(self, x):
             )
 
         h_layer = x_in
-        for layer in self._layers:
+
+        for layer in self._layers[: (2 * len(self.layer_sizes))]:
+            h_layer = layer(h_layer)
+
+        feature_layer = h_layer
+
+        for layer in self._layers[(2 * len(self.layer_sizes)):]:
             if isinstance(layer, APPNPPropagationLayer):
                 h_layer = layer([h_layer, feature_layer, out_indices] + Ainput)
-            elif isinstance(layer, Dense):
-                h_layer = layer(h_layer)
-                feature_layer = h_layer
             else:
                 # For other (non-graph) layers only supply the input tensor
                 h_layer = layer(h_layer)
@@ -359,7 +362,7 @@ def node_model(self):
 
     def propagate_model(self, base_model):
         """
-        Propagates a trained kera model to create a node model.
+        Propagates a trained model using personalised PageRank.
         Args:
             base_model (keras Model): trained model with node features as input, predicted classes as output
 
@@ -403,15 +406,15 @@ def propagate_model(self, base_model):
 
         # pass the node features through the base model
         feature_layer = x_t
-        for layer in base_model.layers:
+        for layer in base_model.layers[1:]:
             feature_layer = layer(feature_layer)
 
         h_layer = feature_layer
         # iterate through APPNPPropagation layers
-        for layer in self._layers:
+        for layer in self._layers[(2 * len(self.layer_sizes)) :]:
             if isinstance(layer, APPNPPropagationLayer):
                 h_layer = layer([h_layer, feature_layer, out_indices_t] + Ainput)
-            elif isinstance(layer, Dropout):
+            else:
                 h_layer = layer(h_layer)
 
         x_out = h_layer

stellargraph/version.py

@@ -1,2 +1,2 @@
 # Global version information
-__version__ = "0.8.2"
+__version__ = "0.8.3"