If the function is defined on a finite domain, you can (in theory) find the fastest algorithm for it: http://www.forwardscattering.org/post/13 http://www.forwar

If the function is defined on a finite domain, you can (in theory) find the fastest algorithm for it: http://www.forwardscattering.org/post/13 http://www.forwardscattering.org/post/13 In practice, something like hill climbing will be done, as you said. Actually it's not even hill climbing, since some of the changes will be pessimisations. (as per the original post)