Symmetrical Building Blocks and the Simple Inversion Operator

摘要

In this paper we analyze a Simple Inversion Operator (SIO), showing that a Genetic Algorithm (GA) with a reordering operation is capable of attaining very good performance on problems compliant with the Symmetrical Building Block Paradigm (SBB). Our approach is different from the classical view of inversion benefits, in that it is not concerned with increasing linkage between different genes, but rather aims at the direct discovery of building blocks with a specific symmetry. Inversion together with crossover works to properly discover and align the building blocks, to form the optimal solution. The main advantages of our approach are the application with good results of a fairly simple and unpretentious operator, and the clear identification of the class of problems where this operator has the highest potential.