You can probably try both and see which works best for you. It may be
that one is better than the other for certain projects. I haven't really
evaluated which one performs best for particular problems but it seems
there are a number of papers that compare the two approaches.
The steady-state algorithm only changes one or two members of the
population at a time whereas the generational version potentially replaces
the entire population (or most of it if you are using elitism).
In general I'd just go with the generational algorithm unless I had a
reason not to. If I remember correctly, I think that in the Watchmaker
framework at least, the generational implementation makes better use of
multiple CPU cores.
Dan.
--
Daniel Dyer