In
the last post we acknowledged that competent GAs are those that automatically
identify building blocks (BBs) and exchange these BBs without disrupting them.
As a simple example, the compact genetic algorithm (cGA) was presented for competently
solving the trap-n functions. In spite of the good results obtained by cGA in these
environments, this algorithm is way too simple for truly tough problems: the m-concatenated trap-n problems, as depicted
in Figure 1. In these kinds of
problems, deceptive trap-n functions are concatenated into a single, bigger
problem.cGA’s probability vector
representation cannot detect the complicated combinations of BBs so, again, a
new strategy to tackle this challenging environments is required: we need an
order-n probabilistic optimization algorithm (in contrast cGA is of order-1).

Following the initial steps of primeval estimation-distribution algorithms (EDAs), a complex probabilistic model is used to detect BBs (instead of the simple probability vec…

Following the initial steps of primeval estimation-distribution algorithms (EDAs), a complex probabilistic model is used to detect BBs (instead of the simple probability vec…