>>DT: > I would suggest that "Darwinian design" has an appropriate
>>analogy with artificial selection ...>
>>BH: > I guess I don't understand what you mean by Darwinian
>>design.>
>>DT I was responding to a comment from Loren which I thought was
>>related to human engineers using darwinian principles to optimise
>>their designs. Objects being designed are provided with the
>>means to adapt to the environment, and the engineers select those
>>characters which emerge as superior from the testing process.
>
>OK. You're talking about various randomization appraches to design,
>including genetic algorithms. This sort of thing has been done, for a
>number of years, with quite a bit of progress made recently due both to
>improvements in computing power and improvements in the theory. I'm not
>sure the analogy with artificial selection is a good one. Genetic
>algorithms are used to perform very complex optimizations on multimodal
>surfaces with beaucoups constraints. Sometimes suboptimal solutions are
>known, and frequently genetic algorithms break into entirely different
>regions in the parameter space, giving unexpected improvements. You really
>do get beneficial "mutations".
It should be added (emphatically) that in no way do I believe the genetic
algorithm or other optimization algorithm is actually "doing" the design.
It's just serving as an assistant to the engineer in locating points in the
design space which satisfy the constraints and optimality conditions. By
virtue of his having specified the constraints and the design criteria, and
perhaps (likely in my experience) having chosen a suitable point from which
to initiatye the search, I would attribute the design to the engineer.
Bill Hamilton | Vehicle Systems Research
GM R&D Center | Warren, MI 48090-9055
810 986 1474 (voice) | 810 986 3003 (FAX)