I created a genetic algorithms dll which, as a parametrized class, accepcts all integer datatypes (but not binary) - and has a few features which go beyond the classical understanding of GA. - If there''s common interest I might make a general dll out of it. A demo donwload is availabe under: http://www.superfluxus.de/EvolutionNo9.zip

quote: Original post by MB
has a few features which go beyond the classical understanding of GA.

Would you care to elaborate on this please? I''d be interested to know what new things you have come up with!

Cheers,

Timkin

Hi Timkin, I just happened a while ago to visit your site, so I know a little bit about it. - My basic concept towards GA is to treat the biologic aspect as pure metaphore and to think of symbolic reprodcution instead (incestuous signs, so to speak). Maybe that's because I've written a lot about cultural theory and machines, and have always noted that the "biological" part is a relative late offspring. You won't find eveolution theory in Babbages work for example. - But to be concrete. Why should I reduce myself on one point, two point, n-point crossover. Why not talk about semantics? - This is exactly what happens when you leave the binary level. You evaluate data and you think about reproduction choices which are different.
To give a speking example. Instead of a mechanical crossover you could look for something as love marriage. - This brings (what the 18th century wanted in inventing "love") some chaotic aspect into the social reproduction system, mixing it up, bringing some very improbable elements into it. For example: one color (species) is looking for a compelementary one. And if you have a deviation vector wghich records the difference to mainstream, you would have a socal convention which could serve as a matrix. Instead of two parental vectors I like to think of two gender
vectors, one male, one female. You may think about a lot of coupling drives, which are very interesting. And they have the wanted effect the mechanics of crossovers point does not have. If colorizes the population, brings in new elements. - This was the reason for me to write a strongly parametrized class, that accepts all kind of data - except binary.

Best martin


[edited by - MB on October 7, 2002 5:07:06 AM]

[edited by - MB on October 7, 2002 5:07:29 AM]

Try looking for some academic papers that have already been researched on the subject (there are a lot). Before you begin making improvements in the field, you should see if you are just repeating previous work.

Thanks for the advice. But that's what I have actually done a while ago (Penrose, Holland etc.) And it was really fruitful, I have to admit. Nevertheless, what I did not found (and the reason for me to get into that field) was something with is not stuck in the natural philosophy dilemma. But maybe I have missed something.

[edited by - MB on October 7, 2002 5:59:10 AM]

You might be interested then in some of the work done here at Monash in experimental ethics and evolving emotions. Check out Kevin Korb''s web page and also the pages of several students (Lucas Hope and Steven Mascaro).

Cheers,

Timkin