Artificial Evolution

[Dr. Hawkins]

On Mon, Oct 17, 2016 at 8:29 AM, Bob Sneidar <bobsneidar at iotecdigital.com>
wrote:

> Some might take this to be a minor point, but Darwinian Evolution relies
> on truely random point mutations followed by the process of natural
> selection. In this case a computer induces a change then evaluates the
> effects of that change. I would rather called this automated engineering.
>

I used to work with Genetic Algorithms and knew somme of the leading
figures.

A GA doesn't evaluate a trial solution so much as applies the solution to
generate a fitness score.  Higher scores are either more likely or
guaranteed a shot at reproducing in the next round.  It's the whole genetic
code rather than the mutation, that reproduces.


> Also, I wonder how many times in Darwinial evolution a species evolved a
> truely quantum leap in it's genetics, only to be wiped out by a flood or
> volcanic explosion, os simply got eaten by it's predator. :-)
>
> You get real world opportunism.

I had a parameter backwards, and the algorithm stabilized at a negative
price in an economic simulation--which was, in fact, the correct but
unanticipated result.

Ashlock was evolving simple code, and put in a random result as the penalty
for not reaching the end of the decision tree--and they adapted to use that
"hazard" as a synchronization mechanism to boost their scores. . .
 (repeated prisoner's dolman type situation)

-- 
Dr. Richard E. Hawkins, Esq.
(702) 508-8462