Re: Oklahoma School Book Selections

[brian whatcott <inet@INTELLISYS.NET>]

At 10:10 11/11/99 -0500, Ludwik wrote:
> Can you provide...
> a verbal description but preferably in Basic.... Something very
> simple and very short.
>
> brian whatcott wrote:
> <snip> <snap>
>
> >  The coding associated with computer implementations is exceptionally
> > straight-forward. Less than 60 lines of code in any language can provide
> >  a useful Darwinian design functionality which handles real valued
> > parameters.
> >
> >    Search for web references for DE  (Differential Evolution) for
> > discussions and examples.


I started by transcribing the Pascal (or was it C?) example on Dorn's
 web page here:

 http://www.icsi.berkeley.edu/~storn/code.html

This example was discussed in the unlikely environs of Dr Dobbs,
2 or three years ago. The article was remarkable for its clarity.

But I see Rainer also has a java example, Mathlab etc., etc.
I also transcribed it to Fortran

Carroll has a worthy Fortran example at

 http://www.staff.uiuc.edu/~carroll/ga.html

Take an input 2-D array of real numbers. This is the list of parameters
 you need to optimize
The array will make room for a population each containing a list
of the parameter set (genes)

Write an objective formula to evaluate the parameter set for an
individual.
 Compose it in the form of a cost function. (Inverse of a merit rating)

Randomize the initial sets of genes in the population.

For each 'subject', split his parameter set with a randomly chosen
 'partner'
Evaluate the cost function of this new parameter set. If this
'offspring' is lower cost than the subject place him in the next
generation array.
If not lower cost, clone the parent subject, and place him in the next
generation to live again.
 During the gene-share, randomize a set proportion of the genes.

The process is iterated until the cost criterion is low and stable.

This is as much as I can briefly explain.
The devil is in the details.

Sincerely

P.S I read that William J. Bennett of The Textbook League
Was able to find two references in print to this OK 'disclaimer'
 in today's editions which he says is modelled upon the
Alabama prototype.

brian whatcott <inet@intellisys.net>
Altus OK