Skip to content
Yi Zhu edited this page May 14, 2021 · 3 revisions

Dense, also known as the full-connected layer.

Dense(;input_shape::Tuple, layer_size::Int64, activation_function::Module, randomization::Bool=true)

input_shape: a tuple contains the total number of inputs, can be set automatically, unless it is the first layer in the network

layer_size: the total number of neurons in the layer

activation_function: a chosen activation function

randomization: generate the weights and biases with Glorot Uniform or set as zeros