| To choose the population size, in the Settings Panel -> General Settings enter the
Number of Chromosomes. The default in GeneXproTools 4.0 is 30, that
is, a very small population, as it produces very good results for virtually all
problems, even for such high-dimensional problems as DNA microarrays
with more than 25,000 variables. But, obviously, if you'd rather
work with larger populations, the limit is 4096 chromosomes in GeneXproTools 4.0.
Learning and adaptation occur in evolutionary algorithms because populations of individuals (models) adapt with time. Obviously, the larger the population the faster
(in terms of generations) adaptation. But there’s a caveat here: in a computer, the higher the number of models the longer it takes to process them. So one must choose a population size that enables an efficient adaptation in record time. In
Gene Expression
Programming, the learning algorithm used by GeneXproTools, this can be accomplished
well enough with populations of 30-200 individuals provided that the right degree of
genetic modification is being introduced in the evolving populations. How this is accomplished is explained in the next section. |
