On Heuristics for Seeding the Initial Population of Cartesian Genetic Programming Applied to Combinational Logic Circuits

The design of circuits is an important research field and the corresponding optimization problems are complex and computationally expensive. Here, a Cartesian Genetic Programming (CGP) technique was used to design combinational logic circuits. Several configurations were tested for seeding the initial population. First, the number of rows, columns, and levels-back were varied. In addition, the initial population was generated using only NAND gates. These configurations were compared with results from the literature in four benchmark circuits, where in all instances it was possible to find that some seeding configurations contributed beneficially to the evolutionary process, allowing CGP to find a solution employing a lower number of fitness evaluations. Finally, the variation of the number of nodes of the individuals during the search was also analyzed and the results showed that there is a correlation between the topology of the initial population and the region of the search space which is explored.