full_layer_type

full_layer_type(
   num_outputs,
   num_inputs,
   num_addit_inputs,
   batch_size,
   activation_function="none",
   activation_scale=1.0,
   kernel_initialiser=empty,
   bias_initialiser=empty
)

The full_layer_type derived type provides a fully-connected (aka dense) layer. The layer contains num_inputs number of input features.

This layer creates a fully- (densely-) connected layer, a standard neural network layer.

Arguments

• num_outputs: Positive integer. Number of output neurons (dimensionality of output space).
• num_inputs: Integer. Number of input neurons. Defaults to number of outputs of previous layer.
• num_addit_inputs: Positive integer. Number of additional inputs to this layer that have been exempt from entering previous layers.
• batch_size: Integer. The number samples in a batch. This is optional (the enclosing network structure can handle it instead).
• activation_function: Activation function for the layer (see Activation Functions).
• activation_scale: A real scalar. Defaults to 1.0.
• kernel_initialiser: Initialiser for the kernel weights (see Initialisers).
• bias_initialiser: Initialiser for the biases (see Initialisers).