Question about how to train a neural network using Flux.jl

Specific Domains Machine Learning
,
2

loss(x,y) needs to return a scalar:

using Flux
model = Flux.Chain(
    Dense(14*16+4, 64, relu),
    Dense(64, 16, relu),
    Dense(16, 1, relu));
x = rand(Bool, 14*16+4)
y = 100
loss(x,y) = sum((model(x) .- y).^2)
opt = ADAM(params(model))
Flux.@epochs 100 Flux.train!(loss, [(x,y)], opt)
2 Likes