Closed acse-ncv19 closed 4 years ago
fc(train_input, config, θ)
doesn't need a label. You can regard θ
as its label if you wish. The existence of labels is to mark different neural networks, but since you provide θ
to the neural network, the neural network is uniquely defined by θ
.
thank you very much for this answer. This is very helpful to know.
config = [20, 20, 20, 20, 20, 20, 20, 20, 1] θ = Variable(fc_init([2; config]))
u_nn = squeeze(fc(train_input, config, θ, "concave")) + 1
I got this kind of code but it only runs if i choose between labling the NN of initialising the weights. When doing both I get this error:
ERROR: LoadError: MethodError: no method matching ae(::PyCall.PyObject, ::Array{Int64,1}, ::PyCall.PyObject, ::String)