gcomgraphfrom2param

Takes a 5d Tensor of size (-1,gs,gs,param,n) and transforms it into (-1,gs,gs,c,c). So basically creates a Graph of size c for each set of n param vectors, using a simple Dense Layer to do the (param,n) to (c,c) transformation.

Arguments

  • gs: The Number of Nodes in each global Graph.
  • param: The Number of Features from which the graphs will be constructed
  • c=2: The Number of Nodes in each resulting subgraph.
  • n=2: The Number of Feature Vectors that result in each subgraph
  • initializer=”glorot_uniform”: The initializer of the Transformation
  • trainable=True: weather the Transformation is trainable