广义回归神经网络预测适应值的人工植物算法

Research on Fitness Approximation in Artificial Plant Optimization Algorithm Using GRNN

人工植物算法是最近几年提出来的一种新颖的智能优化算法,把一个植物的生长过程映射为一个智能优化问题。它为那些高维多模问题提出了一种新的解决方法,但是,当把人工植物算法应用到现实问题中时,有时会遇到适应值很耗时的计算,如优化目标和随机问题中存在随机因素的不确定规划问题,或适应值需要通过很多复杂计算才能近似计算等问题。所以,在人工植物算法中需要采取一些预测适应值的策略,采取了基于代进化控制的结合神经网络预测模型的策略(GAPOA)。

Artificial plant optimization algorithm（APOA） is a recent proposed evolutionary computation in which the growing process of one tree is mapped into the optimized problem.It is suitable to solve the problem of high dimensional multi-modal optimization.One main difficulty in applying APOA to a real-world application is that APOA usually need a large number of fitness evaluations before a satisfying result can be obtained.This paper introduces a new ＂training GRNN using APOA＂（GAPOA） that does not evaluate all new branch positions owning a fitness by the real fitness function and associated reliability fitness of each branch of the plant.This paper undertakes generalized regression neural network（GRNN） to approximate fitness in Artificial Plants Optimization Algorithm.