Abstract
This paper proposes the Noise Analysis compact Genetic Algorithm (NAcGA). This algorithm integrates a noise analysis component within a compact structure. This fact makes the proposed algorithm appealing for those real-world applications characterized by the necessity of a high performance optimizer despite severe hardware limitations. The noise analysis component adaptively assigns the amount of fitness evaluations to be performed in order to distinguish two candidate solutions. In this way, it is assured that computational resources are not wasted and the selection of the most promising solution is correctly performed. The noise analysis employed in this algorithm spouses very well the pair-wise comparison logic typical of compact evolutionary algorithms. Numerical results show that the proposed algorithm significantly improves upon the performance, in noisy environments, of the standard compact genetic algorithm. Two implementation variants based on the elitist strategy have been tested in this studies. It is shown that the nonpersistent strategy is more robust to the noise than the persistent one and therefore its implementation seems to be advisable in noisy environments.