Genetic optimization is the least preferred method for designing single hidden layer nets with given inputs and targets (the universal approximator ) because it is too slow.
However, if different combinations of inputs and network architecture are to be considered, GA may be a good way to go.
Searching in the ANSWERS NEWSGROUP, and comp.ai.neural-nets yields
neural genetic 64 132 699
neural ga 39 69 674
Check the ANSWERS references and respond with specific questions.
