AdaBelief

AdaBelief.fit(;model::Any, input_data::Array{Float32}, output_data::Array{Float32}, loss_function::Any, monitor::Any, α::Float64=0.001, epochs::Int64=20, batch::Real=32, β₁::Float64=0.9, β₂::Float64=0.999, ϵ::Float64=1e-8)

model: any sequential 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: learning rate, default 0.001

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

β₁: β₁, default 0.9

β₂: β₂, default 0.999

ϵ: ϵ, default 1e-8