made SortPooling optional

CHANGELOG.rst

@@ -22,3 +22,9 @@ Version 0.3
 - added flatten_signals option to DeepGraphConvolution
 - publish to pypi and github.
 - publish docs.
+
+Version 0.3.1
+===========
+
+- Make SortPooling optional in DeepConvolutional
+- Host docs on github pages

README.md

@@ -35,6 +35,7 @@ Initiated it with the following parameters:
 |`flatten_signals` (default: False) | If `True`, flattens the last 2 dimensions of the output tensor into 1|
 |`attention_heads` (default: None) | If given, then instead of using <a href="https://www.codecogs.com/eqnedit.php?latex=D^{-1}E" target="_blank"><img src="https://latex.codecogs.com/gif.latex?D^{-1}E" title="D^{-1}E" /></a> as the transition matrix inside the graph convolutions, we will use an attention based transition matrix. Utilizing `dgcnn.attention.AttentionMechanism` as the internal attention mechanism. This sets the number of attention heads used.|
 |`attention_units` (default: None) | Also needs to be provided if `attention_heads` is set. This is the size of the internal embedding used by the attention mechanism.|
+|`use_sortpooling` (default: True) | Whether or not to apply sortpooling at the end of the procedure. If False, we will simply return the concatinated graph convolution outputs.|
 
 Thus, if we have non-temporal graph signals with 10 nodes and 5 features each and we would like to apply a DGCNN containing 3 graph convolutions with hidden feature dimensions of 10, 5 and 2 and SortPooling that keeps the 5 most relevant nodes. Then we would run
 

src/dgcnn/components.py

@@ -71,6 +71,12 @@ class DeepGraphConvolution(layers.Layer):
 
             - (default: (None)).
 
+        use_sortpooling (bool):
+
+            - If False, won't apply SortPooling at the end of the procedure.
+
+            - (default: True)
+
     Inputs:
 
         - X (tf.Tensor):
@@ -149,6 +155,7 @@ def __init__(
         attention_heads: int = None,
         attention_units: int = None,
         flatten_signals: bool = False,
+        use_sortpooling: bool = True,
         **kwargs
     ):
         super(DeepGraphConvolution, self).__init__()
@@ -176,6 +183,7 @@ def __init__(
         self.attention_heads = attention_heads
         self.attention_units = attention_units
         self.flatten_signals = flatten_signals
+        self.use_sortpooling = use_sortpooling
 
         # save kwargs to pass them to the graphconv layers
         self.kwargs = kwargs
@@ -193,7 +201,8 @@ def build(self, input_shape):
             self.convolutions.append(layer)
 
         # initiating SortPooling layer
-        self.SortPooling = SortPooling(self.k)
+        if self.use_sortpooling:
+            self.SortPooling = SortPooling(self.k)
 
         # create AttentionMechanism if required
         if self.use_attention:
@@ -234,7 +243,11 @@ def call(self, inputs):
         # concat them to (N x sum(c_i)) signal
         Z = tf.concat(rec_conv_signals, axis=-1)
 
-        # now apply SortPooling
+        # Check if we apply SortPooling
+        if not self.use_sortpooling:
+            return Z
+
+        # apply SortPooling
         Z_pooled = self.SortPooling(Z)
 
         if not self.flatten_signals: