Internal vs. External Parameters in Fitness Functions

A fitness function is needed for a Genetic Algorithm (GA) to work, and it appears natural that the combination of objectives and constraints into a single scalar function using arithmetic operations is appropriate. One problem with this approach, however, is that accurate scalar information must be provided on the range of objectives and constraints, to avoid one of them from dominating the other. One possible solution, then, is to try to join the objectives with the constraints with internal parameters, i.e., information that belongs to the problem itself, thereby avoiding external tuning. The building of the fitness function is so complex that, using internal or external parameters, any optimal point obtained will be a function of the coefficients used to combine objectives and constraints. However, it is possible that using internal parameters will increase performance compare to external ones.