GA - SkyWorld117/DianoiaML.jl GitHub Wiki

GA is Genetic Algorithm.

GA.fit(;models::Array, input_data::Array{Float32}, output_data::Array{Float32}, loss_function::Any, monitor::Any, α::Float64=0.01, gene_pool::Int64, num_copy::Int64, epochs::Int64=20, batch::Real=32, mini_batch::Int64=5)

models: an array of models

input_data: a 2-dimensional input data in a shape of (,batch_size)

output_data: a 2-dimensional output data in a shape of (,batch_size)

loss_function: a loss function

monitor: a monitor

α: mutation rate, default 0.01

gene_pool: the number of models

num_copy: the number of models that will be copied into the next round directly

epochs: number of training epochs, default 20

batch: the number of batches for each training epoch, default 32

mini_batch: the size of minibatch for each update, default 5