Inverse design using Matlab ANN

I am trying to write a code using built in function of ANN in Matlab in order to obtain a multi-variable solution for a given output. Let me clarify.

Suppose z=f(x, y,t). I have a set of values of input say x1, y1, t1; x2, y2, t2...etc for which I know the result z1, z2......etc. for each set. I want to use these data to train a buit-in ANN of Matlab. Finally I want to predict possible values of x,y,t for a new given value of z. I understand that there can be multiple set of solutions. I am looking for the best solutions which correzponds to the closest output to z.

I have some idea of solving the similar problem using 'fitnet' and 'train' function in Matlab to solve the above probem for seeking forward solution. I mean, I can predict z for a given new set of (x,y,t) for the similar problem. In this case, there is a single output. But how to solve it for inverse case in which there can be multiple set of solutions.

Any help will be highly appeciated since I am very beginner of using ANN tools in Matlab. Thanks in advance.