Neural Network Setup for Function Maximisation

Hi,

I have a complex, multi-layer neural network, and I wish to maximise a function of the final output from this network.

The output of the network as it stands, is a single value, d. I wish to maximise X(t) = X(t-1) + d(t)*a + b*(d(t) - d(t-1)), where a is an input variable; b a known constant; X(t), d(t) are the values of X, d at time t; X(t-1), d(t-1) are the values of X, d at time t-1.

I also require that X feeds back into other layers of the network.

Is there is a way I can set up a final layer of the network to represent the above function, and then use the out of the box capability to optimise the network? If not, are there any other suggestions on the best way to achieve this please?

Thanks,

David.