Yep, took a population of I think 8 points (stored in matrices, limiting the number available, since the calculator only has 10 available, at least 2 of which needed to be kept empty for use in the selection stage to avoid programming headach or unintentional loss of specimens), ran em through the function provided and used that number to do fitness testing, highest/lowest 3 were caried on to next step. These three were then reproduced with one another; one decimal place was a single gene, with it being programmed for 4 decimals of accuracy, making it 5 total genes. Reproduction involved cloning of the top 3, then mutating those clones, as well as combining pairs of the three w/ mutation. It had elitism as well, keeping a copy of the best solution unmutated.
3 best: A, B, C
after reproduction: A, Am, Bm, Cm, ABm, ACm, BCm, ABCm*
It did have some major problems though brought about by the way it was programmed. Since each decimal was a gene, going from 0.9999 to a correct solution of 1 is improbable; likewise, from .0999 to .1, ect are also improbable. However, I never bothered fixing it since it runs to slowly on a calculator to be of much practical use anyway, except in very rare cases. One generation takes about 1 to 2 seconds to go through, which is why it takes so long to find a solution.
The way I wrote it, the genetic algorithm itself is a massive multi-hundred line code, with the function being tested put into a seperate program for ease of access. Without too much alteration, it could probably be used for other genetic algorithm-y purposes, albiet ones which would take a very long time.
Oh, and I nearly forgot another experiment I carried out with it... You see, it was taking quite a while to come up with the correct answers, as I have stated before. So, I rigged up a much more simple genetic algorithm which would run that genetic algorithm over and over, modifying its mutation rate. Overall, it took about 8 hours to run through it completely, but by the end of running it three times the ideal mutation rate was nearly identical(margin of difference of about 2%), and so I now use that as my mutation rate (about 26% btw, although it would certainly vary from one GA to another).
*threesome on a calculator, bow chicka bow wow