multidense¶
Runs just a list of Dense Layers (defined by m.mdense* and by the parameter q, which gives width and number of Layers) on the last axis of the input data
Arguments
- g: A grap Class containing an Adjacency Matrix and a Feature vector, as well as a state Class containing the current standart Values for gs(=Graph Size, Number of Nodes in the Graph) and param (=Parameter Count, Number of Features for each Node)
- m: The Constants defining the Function Behaviours (matching Parameters below)
- q: An Array setting the size of each Layer [8,4,7] creates 3 Dense Layers with 8,4 and 7 nodes respectively
Constants defined in m
- mdense_activation=”relu”: The Activation of each Dense Layer
- mdense_init_kernel=tf.keras.initializers.Identity(): Kernel Initializer of each Layer
- mdense_init_bias=tf.keras.initializers.Zeros(): Bias Initializer of each Layer
- mdense_usebias=True: Should this Networks use a Bias Term
- mdense_batchnorm=False: Should you use a BatchNormalisationLayer after each Layer