Greetings.

How can I constrain a multilayer feedforward neural network so that
the approximated function will be strictly monotonous? Basically I
know that the function to be approximated is strictly monotonous and I
would like to use this hint during training. Currently due to my wierd
objective function network finds a better function that minimizes my
objective function, but my problem is the function assumed by the
network is not monotonous, I can actually see this by plotting the
network input output pairs after training. I need a method to
constrain the network so that it will only search for monotonous
functions while minimizing the objective function.

Thanks in advance.