科研之友
微信
新浪微博
Facebook
分享链接
摘要
Many engineering applications can be approached as optimization problems whose solution commonly involves the execution of computational expensive objective functions. Recently, Evolutionary Algorithms ( EAs) are gaining popularity for solving complex problems that are encountered in many disciplines, delivering a more robust and effective way to locate global optima in comparison to classical optimization methods. However, applying EA's to real-world problems demands a large number of function evaluations before delivering a satisfying result. Under such circumstances, several EAs have been adapted to reduce the number of function evaluations by using alternative models to substitute the original objective function. Despite such approaches employ a reduced number of function evaluations, the use of alternative models seriously affects their original EA search capacities and their solution accuracy. Recently, a new evolutionary method called the Adaptive Population with Reduced Evaluations ( APRE) has been proposed to solve several image processing problems. APRE reduces the number of function evaluations through the use of two mechanisms: ( 1) The dynamic adaptation of the population and ( 2) the incorporation of a fitness calculation strategy, which decides when it is feasible to calculate or only estimate new generated individuals. As a result, the approach can substantially reduce the number of function evaluations, yet preserving the good search capabilities of an evolutionary approach. In this paper, the performance of APRE as a global optimization algorithm is presented. In order to illustrate the proficiency and robustness of APRE, it has been compared to other approaches that have been previously conceived to reduce the number of function evaluations. The comparison examines several standard benchmark functions, which are commonly considered within the EA field. Conducted simulations have confirmed that the proposed method achieves the best balance over its counterparts, in terms of the number of function evaluations and the solution accuracy.
- 出版日期2016
