gcomdepoollg¶
Extension of gcomdepool by also learning a Graph. Breaks Graph Permutation Symmetry, by using a Dense layer to create a c*c Adjacency Matrix from all features. Returns the Feature Vectors and the Adjacency Matrices
Arguments
- gs: The initial Number of Nodes of the Graph (Graph Size)
- param: The initial Number of Features for each node
- paramo: The Number of Features each Outputvector should have
- c=2: The Factor by which to increase gs
- metrik_init=glorot_uniform: The Transformation initializer to go from little Feature Vector to output Vectors
- graph_init=keras.initializers.Identity(): The Initializer to create the Graph
- learnable=True: Shall the Transformation be learnable