Re: [asa] Parameterized Evolution

This is an interesting and valuable discussion, I think. Could some of you comment on how this applies to specific observations? I'm particularly thinking of the all-too-common arguments along the lines of "there must be a designer because abc is so well tuned" or "there can't be a designer because xyz isn't optimized." Neither path seems reasonable to me but these arguments continue. There are many specific examples but I'm especially intrigued by the trade-off examples. For instance:

1. genetic anomaly in which having one copy protects from malaria but two copies leads to sickle cell anemia
2. juxtaposition of the larynx and the pharynx which occurs uniquely in humans, enabling speech on one hand but inducing susceptibility to choking on the other
3. relative position of the retinal cells and optical nerves in the human eye

Are these examples of local minima as opposed to optimized states? What are some others?

It is often assumed in discussions that a particular feature in a species exists because it has proven to be successful through a long period of evolutionary development. How can one tell whether a feature is a local minimum or an optimized state? Does it really matter?

Randy

Iain wrote:

    I am actually quite sceptical about the ability of genetic algorithms to avoid local minima. They often exhibit a phenomenon called "premature convergence" where the entire population is clustered around the local minimum. Now although the GA has the ability to make large jumps via crossovers, in a high dimensional space it is extremely unlikely that a big jump will increase the fitness. This is due to a well known problem called "the curse of dimension" ie the exponential increase in the volume of the search space with dimension.

To unsubscribe, send a message to majordomo@calvin.edu with
"unsubscribe asa" (no quotes) as the body of the message.
Received on Mon Aug 25 10:12:41 2008