Too simple for Galapagos?

Hi,

I am trying to understand the essentials of Galapagos and got stuck with a simple task: find a number that is most similar to number 5 (see attached definition).
Galapagos shows in it's genome window very quickly that all of the genomes are 5. But after that it is impossible to stop the solver except of switching to the "Options" or "Record" tab. When switching to the "record tab", the solver re-initiates and runs for a quite long time.
In any case, the Galapagos window freezes for a certain time and the number 5 comes never up in the genome slider. Even when reinstating a genome that is 5, the numbers that come up in the genome slider are randomized through the whole range I have defined.
Is this a wrong example for Galapagos or is something wrong with my way of defining the task or understanding the results?
Thanks for any comment or explanation.

Attachments:

galapagos test.ghx, 14 KB

Replies to This Discussion

Permalink Reply by David Rutten on October 18, 2010 at 3:52am

Hi Hrvoje,

in your file, the fitness is always 5, no matter what the value of the slider. So one slider position doesn't result in a better fitness value than any other slider position, which in turn means Galapagos has got nothing to go on.

Your fitness value needs to reflect the fitness of the slider position. Then, Galapagos needs to be told what it is supposed to be looking for. I.e. it should try and minimize the fitness value (lower = better), maximize the fitness value (higher equals better) or optimize the fitness value (closer to X = better).

I attached two files, one extremely simple, one a bit more involved. They both solve the same problem. Be sure to have a good look at the Galapagos settings, as they differ in both files.

There's one more thing I should say. Galapagos has algorithms build in that prevent the same genome from occurring more than once in any given species. If you have an Integer slider with 11 possible settings, that means you can only have 11 unique genomes. Galapagos will exhaust this pool in half no time, after which it will seem to jump randomly back and forth, desperately trying to find another unique genome (and always failing). This is why I used floating point sliders, to increase the number of possible unique genomes.

--
David Rutten
david@mcneel.com
Poprad, Slovakia

Attachments:

galapagostest.ghx, 24 KB
galapagostest_simplest.ghx, 14 KB

Permalink Reply by Hrvoje Petrovic on October 18, 2010 at 4:20am

Hi David,

Thanks a lot for your time and the explanation. I am now getting closer with my practical understanding of the essentials. I have been through your wonderful and extensive blog post about evolutionary principles, but it is sometimes hard to understand which buttons to push and why to push them...

Cheers,

Hrvoje.

RSS