Evolutionary GMDH-based Identification of Building Blocks for Binary-Coded Systems

This paper proposes an approach to the problem of building block extraction in the context of evolutionary algorithms (with binary strings). The method is based upon the construction of a GMDH neural network model of a population of promising solutions with the aim of extracting building blocks from the resultant network. The operation of the proposed method is regardless of the order by which building blocks are positioned in strings representing the solutions. The experiments are carried out on some well-known benchmark functions including DeJong´s