How to use BINARY Cross Entropy function in a dlnet

I tried to calculate a binary cross-entropy by

perf = crossentropy(net,targets,outputs,perfWeights)

Yet it seems it supports a normal net only, not a dlnet for custom training.

Plus, if I just stacked up layers of a net without having trained it, how can I convert a layer object into a neural network object? Is it possible?

Futhermore, there is a CATEGORICAL cross-entropy function supporting dlnet, but that is not what I want

dlY = crossentropy(dlX,targets)

Thanks.