In order to test speed of convergence, the new operator was employed to optimise some well known functions (outlined in
Yao [11]). Example times for convergence in iterations and seconds are given,
together with some discussion where relevant. The
evolutionary strategy he used 
-ES is not directly mappable to our scheme, instead, we
use a population of 200 (except where stated) with replacement as outlined in section I.
Note that for unimodal functions, the issues of speed of convergence as well as quality of solution are important for our applications. For polymodal solutions, any equally important solution is useful. We do not claim that we are able to find either all solutions or the global optimum. For a discussion of how we can find all optima, see section V.
In all cases, we give the average results over ten runs for clear convergence to an optimum, in iterations (i.e no further improvement). Note that we do not give timings as all experiments took less than 5 seconds on a 1GHz Athlon PC system running the Linux/GNU operating system.
We also give graphs showing averaged fitness function values of the best chromosome in the Appendix. Note that the fitness function numbering is as in Yao [11].