>3) One of the most useful situations for writing Evolution-style computer
>programs is when standard approaches get stuck in local optima. Why?
>Evolutionary-style programs do not get stuck in local-optima. I refer
>you to "An Introduction On Genetic Algorithms". It's written by the
>brilliant Melanie Mitchell, published by MIT Press and is available at
>the low cost of $30.00.
Genetic algorithms most certainly do get stuck in local optima for the
same reasons you mention below in reference to Hillis's work and also
because evolution requires selection pressure. Some species haven't
changed in a billion years because there has been no pressure to evolve.
>4) There are exceptions to 3). For example: W. Daniel Hillis and his
>experiments with using GAs to discover sorting networks. However, his
>experiments had a limit on 1) The number of generations allowed, 2) The
>time allowed for the total experiment and 3) The processing power
>available to throw at the problem. Real world Evolution suffers none
>of these limitations.
Last time I checked the real world had a finite number of organisms
and number of generations.
BTW, for everyone else who found replies to Mr. Leeper's mail without
seeing the original, your mail is probably sorted by date. Apparently
Mr. Leeper is communicating with us from last month.
-- David McFadzean david@lucifer.com Memetic Engineer http://www.lucifer.com/~david/ Church of Virus http://www.lucifer.com/virus/