Neural Network non mutually exclusive cross entropy loss function

Hello,

I have a neural network with 100 outputs which solves a binary classification problem, ie every output can take value 0 or 1. I want to use cross-entropy performance function but as I see it only supports classification with mutually exclusive outputs. Is there any way to use crossentropy for non mutually exclusive outputs? I have found this answer but it seems not to work. ( https://www.mathworks.com/matlabcentral/answers/213915-can-target-vector-of-neural-network-have-sum-of-columns-1 ) Thanks in advance.

Answer 1

[Greg Heath]

The target matrix columns should contain the

        NON-NEGATIVE UNIT SUM APRIORI CLASS PROBABILITIES !!!

See any text regarding statistical classification

Hope this helps

Thank you for formally accepting my answer

Greg