Gradient decent on the inputs of a pre-trained neural network to achieve a target y-value

I have a trained neural network which suitably maps my inputs to my outputs. Is it then possible to specify a desired y output and then use a gradient decent method to determine the optimum input values to get that output?

When using backpropegation, the partial derivative of a weight is used with error function to proportionally adjust the weights; is there a way to do something similar with the input values themselves and a target y value?

[Greg Heath]

So far, I don't know what you are doing or why. What are the dimensions of your input(I-dimensional x) and output (O-dimensional output y and target t) vectors? How many examples(N) and how many hidden nodes()?

Can you explain the problem in terms of the I-H-O network and what you want as a final answer?

desired answer = ?