Target Data Contains NaN values

Hello,

I'm having issues dealing with NaN values in neural network training and simulation. The input data set is a 24x12 matrix with no NaNs but the output data set is a 15x12 matrix with several NaNs. After training, the simulation results return numbers for the NaNs. Please is there a way I can program the network for the NaNs not to be replaced by numbers after simulation ? I've attached the Input data (I.mat) and output data (O (2).mat).

Thank you for any help.