Generating extrema approximation of analytically incomputable functions through usage of parallel computer aided genetic algorithms

TL;DR

This paper demonstrates how parallel genetic algorithms can efficiently approximate extrema of analytically incomputable functions, with significant speedups on modern multi-core processors.

Contribution

It introduces a parallelization of genetic algorithms using OpenMP for better approximation of complex functions and compares various genetic operator modifications.

Findings

Significant speedup on multi-core processors using OpenMP

Different genetic operator modifications affect solution evolution

Parallel genetic algorithms effectively approximate extrema of complex functions

Abstract

This paper presents capabilities of using genetic algorithms to find approximations of function extrema, which cannot be found using analytic ways. To enhance effectiveness of calculations, algorithm has been parallelized using OpenMP library. We gained much increase in speed on platforms using multithreaded processors with shared memory free access. During analysis we used different modifications of genetic operator, using them we obtained varied evolution process of potential solutions. Results allow to choose best methods among many applied in genetic algorithms and observation of acceleration on Yorkfield, Bloomfield, Westmere-EX and most recent Sandy Bridge cores.