Next: Model Parameter Encoding
Up: Standard Genetic Algorithm
Previous: Standard Genetic Algorithm
The first step, of course, is to select the model parameters related
to the ``physics'' of the problem, their ranges (maximum and minimum values)
and their required resolution. Then it is time to choose the evolution
parameters actually related to the genetic algorithm itself. The performance
of a genetic algorithm to solve a particular optimization problem depends
critically on the choice of its evolution parameters that must be fine-tuned
to that problem as much as
possible. In general it is difficult to give hard and fast rules that may work
with a wide range of applications, although some guidelines exist
Goldberg and Deb (1991); Goldberg and Richardson (1987); Goldberg (1989b). For this problem, I choose the
evolution parameters one by one
starting with the most critical and working my way down to the least critical.
Once a particular parameter is selected it is kept constant in the tests
to select the remaining parameters. I do so because it would be nearly
impossible to test all possible combinations. The reader may find useful to
refer to Appendix A for a description of the evolution parameters themselves.
Next: Model Parameter Encoding
Up: Standard Genetic Algorithm
Previous: Standard Genetic Algorithm
Stanford Exploration Project
11/11/2002