[galib] Dynmatic cast problem?
hneil at fastwebnet.it
Fri Oct 14 13:00:18 EDT 2005
I'm pleased to be a part of this mailing group. I hope you don't mind me
mailing this. I thought one of you might be interested in a thesis topic
The idea is simple. Genetic algorithms are wonderful when it comes to tweaking
variables to finding the right mix towards achieving a specific objective.
However, there are many variables associated with genetic algorithms that
must be tinkered and tweaked (elitism, mutation method, crossover method,
So my idea is this: Instead of having an enormous population of 1000 individuals,
why not split them up into 10 genetic algorithms of 100 individuals each.
There would be a "higher order" genetic algorithm which would tweak the
settings of the lesser genetic algorithms to ensure that each is progressing
towards the objective at a reasonable rate. For this, the fitness would
be a combination of the best individual from the population and the rate
of improvement (we will only replace those ones which fail to show a good
individual or good improvement).
What do you think?
Neil S. Harmon
hneil at clemson.edu
More information about the galib