Abstract
We investigate the distribution of performance of the
parity and always-on Boolean functions given only the
appropriate building block. These problems have
``needle in a haystack'' fitness landscape and so are
unsuitable for genetic programming or other progressive
search techniques. Theoretical analysis shows in the
limit as program size grows the density of solutions is
independent of size but falls exponentially with number
of inputs.
Users
Please
log in to take part in the discussion (add own reviews or comments).