Evolution Algorithms

[David Kyle]

I was hoping someone here might have some information for me. I’m
interested in how start coding programs that would “evolve”. Ultimately
I thought I might trying to build a program that would evolve and grow
based on the outcomes of tic tac toe games. Maybe even step it up to
checkers later on. I’m a beginner to this topic area however I’m an
experienced programmer.

I was hoping to get some referrals to books or websites that might be
able to help me get started.

Why do I want to do this? Well just as an interests sake. There are so
many things out there that try to simulate intelligence. Maybe we
shouldn’t try to simulate it? Maybe we should just grow it? I don’t
know, it just seems like a new science to me that has just recently
become available to practice because of all the computer power it needs.
Something like fractals. I would just like to explore it my self for a
bit.

Any comments are welcome too I’m interested in anyone’s thoughts on this
subject.

 

[Bruno Jouhier [MVP]]

One of the best ones I read recently: "Blondie 24" by David Fogel.
It brilliantly demonstrates how a darwinian process can produce a master
level checkers player.
http://www.amazon.com/exec/obidos/t...f=sr_1_1/002-8316725-4042458?v=glance&s=books

You can also take a look at the "Creation" book by Steve Grand. Link is on
the amazon page above.

Bruno.

 

[David Kyle]

Thanks for the reference to the book blondie 24. I looked over it and
will be purchasing it shortly. Looks like someone beat me to the punch
with the checkers idea but that's not surpising. I've always deamed
that programming is largely just researching solutions already built by
someone else. The rest is just adapting it to your needs.

For once though I would like to be pushing the leading edge instead of
just keeping up with it if you know what I mean?

Thanks again!

 

[Bruno Jouhier [MVP]]

Thanks for the reference to the book blondie 24. I looked over it and
will be purchasing it shortly. Looks like someone beat me to the punch
with the checkers idea but that's not surpising. I've always deamed
that programming is largely just researching solutions already built by
someone else. The rest is just adapting it to your needs.

Yes!

Getting back to what this NG is about, I wrote (for fun, after reading the
Blondie book) a small C# program that solves the traveling salesman pb with
a similar approach. It starts with a random path and mutates it and selects
the best candidates for the next generation. Works quite well (you get an
excellent solution on 100 points with a few thousand iterations). I wrote a
sketchy WinForms interface that lets you view how the solution converges. If
you're interested, I can send it to you.

Bruno.