bvanaken/clinical-outcome-prediction

Question: Mortality Prediction having Two Output Neurons

Closed this issue · 1 comments

Hi,

I was implementing mortality prediction while doing analysis I come to know that the final layer has two output neurons,

classifier): Linear(in_features=768, out_features=2, bias=True)

Generally in the case of binary classification problems we use only use one neuron with sigmoid right?

Hi,

we used the FARM framework for training our models. The standard way of treating binary classification in FARM is to use two output neurons, probably because of simplicity. Anyway, both approaches should yield similar results, so you could also use one neuron with a sigmoid.