Another important aspect of our approach is the extent to Which we leverage expert knowledge .A desirable property of genetic algorithms is the extent to which mechanism is can be domain-independent. In the interest of improving the performance of our algorithm we apply domain expert knowledge in several ways. The initial pool does not consist of randomly generated chromosomes, but rather chromosomes with uniform values selected from the set of allowed values. This is closer to the way an expert designer would begin working on a problem. Most of the operators we apply are based on design practices described by domain experts, although the decision of when to apply those operators, and the evaluation of the result, is handled entirely by the evolutionary nature of the algorithm. We do not include random mutation, but instead apply of some of the domain expert operators to a randomly-selected fraction of the solution members. We do include a crossover operator, but it turns out to play a relatively minor role.