GA for deceptive problems: inverting schemata by a statistical approach

We propose a method to overcome the premature stagnation of genetic algorithms (GA). We first define an additional population in order to store a larger part of the resulting chromosomes. We then extract the schema responsible for the stagnation of the algorithm, derive its complementary schema and resume the GA´s evolution with some fixed positions in the chromosome. We prove that, if working with a directed mutation, the GA will explore better than if it carried on with the canonical genetic operators