class

ML::Layer

Inherits Reference < Object

Single layer in neural network

Constructors

new(input_size : Int32, output_size : Int32)

backward_delta(output_delta : Array(Float64)) : Array(Float64)

Compute delta for previous layer

biases

forward(input : Array(Float64)) : Array(Float64)

Forward pass

get_weights

Get weights (for saving)

set_weights(weights : Array(Array(Float64)))

Set weights (for loading)

update(input : Array(Float64), delta : Array(Float64), learning_rate : Float64)

Update weights and biases

weights