Parallel Genetic Algorithms
Ron Shonkwiler · 1993
A universal method for parallelizing a Genetic Algorithm is given. Referred to as IIP parallel, independent and identical processing, theoretical analysis shows that the technique achieves a speedup using m processors given bymsm−1 where the acceleration factor s is a parameter depending on the details of the GA. Typically s > 1. The results are illustrated on two problems small enough that the exact calculation of s can be made and compared to the asymptotic estimates. The results are also illustrated on the non–trivial Inverse Fractal Problem. It is noted that the same results have been attained elsewhere on a wide variety of well known significant problems.