Why are Genetic Programming parameters such as "probability of
mutation" constant (i.e. these parameters don't depend on  generation
number, quality of the best solution to this point, etc)?  What is it
about Genetic Programming that sets it apart from algorithms such as
simulated annealing where the counter part of "probability of
mutation" parameter changes from one iteration to the next (i.e.
probability of accepting a solution that is worse than the best
solution thus far is gradually reduced in simulated annealing).  [Does
the reason for the above have to do with the fact that GP searches
over a space of functions and simulated annealing searches for a
numerical solution?  Why does it make any difference]

Please do not provide links, or point out that this is the setup of
most GP studies.  My question is basically "Is there a reason why this
was chosen as the setup?"

Thank you for your help!

---

Request for Question Clarification by pinkfreud-ga on 14 Mar 2006 10:11 PST

This is outside my areas of expertise, but I am puzzled by your
requirement that we not provide links. Are you seeking someone's
unsupported opinion on this?