摘要
在许多优化问题中,寻找最优解并不是唯一目的,更重要的目标往往是“进步”,算法的优化作用更在于维持一个稳定的改进过程。本文从这一角度出发进行了对传统遗传算法的改进,强化了其渐进收敛和进化能力;并考虑与反向传播学习算法的有机结合以提高训练精度及泛化能力。以一类解析型模糊系统的建模过程为例进行的仿真研究表明,新算法更适用于这类优化问题,能够在更短的时间内有效地改善系统性能;同时,具有很强的自适应性,应用简便也是它的主要优点之一。
For many practical optimization problems, to find the optimal solution is not the only motive. The more important target may be to gain advancement, i.e. to make steady and efficient progress on running performance. From this point, this paper presents a new improving mode of Genetic Algorithm (GA). By emphasizing gradually evolutionary abilities of GA, it maintains both better quality and shorter time of convergence. Moreover, the combination with Back Propagation (BP) learning method may advance precision and generalization of the result. Simulation results of its application on a kind of analytical fuzzy system modeling show that it is more suitable for optimization problems, it can achieve favorable performance in less time. In the meantime, being more adaptive and convenient in use is also one of its advantages.
出处
《系统仿真学报》
CAS
CSCD
2001年第4期494-497,共4页
Journal of System Simulation