News for hackers. Stuff that matters.
Re:Evolutionary algorithms and analog electronic circuits by ycui1986 on Sunday July 27, 2014 @02:25AM (#8092138)
This kind of work was published on scientific american about ~10 years ago.

Anyway this post was more detailed than the introductory on scientific america. This kind idea did not expolode over the past ten years. Something in detail might be wrong.

Re:Evolutionary algorithms and analog electronic circuits by Iv on Sunday July 27, 2014 @03:18AM (#8092202)
Most genetic algorithms break when you try to make them solve too complicated problems. They do not scale really well. Mother Nature has actually a lot of hacks to scale well. Evolution is a process that evolved by itself. If you want a genetic algorithm to solve complex problems, at one point you have to let the reproduction algorithm itself to evolve.

To give you an idea, only recently have we discovered that totally different species can exchange genetic sequences through virus (it is called horizontal gene transfer) and we are not really sure why sexual reproduction is a good idea when you can already mutate, have horizontal transfers and clone yourself.

"Make random mutations, select the best" is really not how nature works.

Re:Evolutionary algorithms and analog electronic circuits by paulbaumgart on Sunday July 27, 2014 @03:48AM (#8092233)
Interesting. Do you have any good links for learning more about the evolution of evolution? Thanks!
Re:Evolutionary algorithms and analog electronic circuits by csirac2 on Sunday July 27, 2014 @05:48AM (#8092350)
It's interesting to read about hypothesized age of plant (and I guess animal) species based on their molecular "clock rate" (rate of mutations) in different parts of the genome.

The mutations indeed are far from random. More important regions which fundamentally affect the viability of the organism evolve incredibly slowly and conservatively. The less important bits are less restricted and mutate more readily.

Additionally it's interesting to consider an individual in the context of its lineage. Some branches are extremely rich in diversity but mightn't have the population numbers of more "boring" clades.

It's easy to mistake success at population size as the end-game, the winner if you will, of evolution. Population size helps but at the end of the day it might be the smaller population, more diverse clades that are more adaptable to environment changes and survivability into the future in general.

Re:Evolutionary algorithms and analog electronic circuits by Houshalter on Sunday July 27, 2014 @06:38AM (#8092421)
Well evolution can evolve things like DNA repair, gene transfer, and sex. That's a very different thing from evolving the representation itself. Even if it did happen, it's not remotely practical for computer simulations.
Re:Evolutionary algorithms and analog electronic circuits by nullc on Sunday July 27, 2014 @07:42AM (#8092505)
Tierra evolved— without extensive setup in the representational structure— viruses and other kinds of interesting parasites and countermeasures, including sexual reproduction: http://infidels.org/library/modern/meta/getalife/coretierra....

(this isn't to say that evolutionary approaches are all that useful, they're usually outclassed by other optimization approaches, and where they work it's either in trivial cases or with mind-boggling amounts of computation)