Mutation rate changing
koolaid2000man at yahoo.com
Mon Oct 21 16:38:14 EDT 2002
i would suggest rethinking that statement about 100%
mutation rate for starters. The mutation rate aims at
bringing diversity to the population once the
population starts converging towards a certain part of
the search space. You should employ crossover
initially to make use of the inherent diversity you
have at the initial populations due to random
initialization at the start till crossover has no more
diversity to employ at which point you need to start
using mutation more heavily to bring diversity in. So
use xover 1st to make use of already existing
diversity then mutation to introduce diversity once
xover no longer has any functionality since it starts
crossing between members of the search space that
whirl around the same area. Trying looking into
Whitley's papers available at the nec computer science
repository and De Jong's thesis at his webpage.
--- Finger <fpaint at newmail.ru> wrote:
> I have some question about changing of the mutation
> rate during GA proccess.
> In a simple GA with a high genome length and a not
> so large population size
> the mutation rate must to be near to 100% at the
> begin of the proccess and
> as small as a zero at the end, otherwise the procces
> will be too long and
> pearhaps will don't find the optimal solution. Where
> I can find some kind of
> algorithm of current mutation rate calculating? I
> guess it will depend on
> the population properties such as fitness variety.
> With best regards, Finger.
> galib mailing list
> galib at mit.edu
Do you Yahoo!?
Y! Web Hosting - Let the expert host your web site
More information about the galib