gcompool

Orders a Feature vector of size (-1,gs,param), orders it by its last Feature and uses a simple Dense Layer to transform each c neighbouring Feature Vectors into one Feature Vector of size paramo (-1,gs/c,paramo)

Arguments

• gs: The initial Number of Nodes in each Graph.
• param: The initial Number of Features in each Node.
• paramo: The Number of Features in each output Node.
• c=2: How many Input Feature Vectors are to be transformed into one Output Vector
• metrik_init=”glorot_uniform”: Initializer of the Transformation
• learnable=True: Weather the Transformation is learnable.