Re: Any Ada Genetic Algorithm?

["Bobby D. Bryant" <bdbryant@mail.utexas.edu>]

On Fri, 01 Mar 2002 23:40:59 -0600, Bobby D. Bryant wrote:

> On Fri, 01 Mar 2002 21:33:24 -0600, Adrian Hoe wrote:
> 
>> I am looking for works on Ada Genetic Algorithm. I searched Google
>> and found the VAGAS is the only work. But it is 7 years old and the
>> links are broken.
>> 
>> Any pointers?
> 
> Just FYI, I'm using Ada to write GAs for my Ph.D. research, and all
> the code will be released under the GPL after I'm done... someday.
> 
> (What I'm doing is actually a narrow subset of GA technology called
> neuroevolution, the use of GAs to generate simulated neural networks
> for controllers, pattern matchers, etc., that can be "trained" by
> sparse reinforcement.)

Bah, I'm having trouble with accidentally hitting the accelerator keys
for my newsreader.  Where was I?

I do have some working code you can download for another rather special
subset of GA, namely a solver that works by reshuffling a single string
of pseudo-DNA rather than breeding two strings.  It comes with a
simulator for applying it to the TSP (and some unimplemented stubs for a
couple of other things).

I suspect it would be difficult to convert for use as a general purpose
GA, so you might find it easier to start from scratch if you're not
interested in the TSP or some similar problem.

You can get the code at
http://www.cs.utexas.edu/users/bdbryant/talk-origins/ga-on-tsp/index.html,
in a link under the section titled "Source Code".  The code is
needlessly complex due to what I was working on at the time, and not
really clean enough for publication, but it should compile and run if
you want to play with it.  It is GPL'd, and presently unmaintained.  It
requires GtkAda because it has an animated display, and it will take a
bit of hacking to make it work if you have upgraded to the latest
version of GtkAda.  (It should compile as is with GtkAda 1.2.11.)

The TSP simulator will generate a random map for you, or you can use it
on problems from the TSPlib; see the README for details about finding
that if you want it.

You can see some screenshots of it at the URL I mentioned above.  You
may find the site rather strange; it was written to support some
discussions on talk.origins, where certain creationists are pushing
William Dembski's bizarre notion that genetic algorithms are no better
than blind searches.  (OK, that's true on *some* problems, and also true
if averaged across *all* possible problems... but the site shows that it
is not true on *every* problem, as Dembski apparently wants his
followers to conclude.)

Due to the odd purpose of the site, the code has been hacked to do a
blind search in parallel with the GA for side-by-side comparison, but
the README tells you how to remove this.

Enjoy,

Bobby Bryant
Austin, Texas