求解多目标优化问题的随机梯度遗传算法

遗传算法的收敛速度很慢，为此引入另一种解决优化问题的工具，即Simultaneous Perturbation Stochastic Approximation(SPSA)算法，该算法是一种简单、易实现、高效率的随机逼近算法。本文将SPSA算法作为一种快速局部优化方法并将其和遗传算法的整体搜索策略结合起来，提出一种解决多目标优化问题的随机梯度遗传算法，对新算法的执行策略进行了认真的设计。大量的数值实验表明：随机梯度遗传算法不仅提高了多目标遗传算法的收敛速度，且得到了大量的分布较均匀的Pareto最优解。

Stochastic Gradient Genetic Algorithm for Solving Multiobjective Optimization Problems

The convergent speed of genetic algorithm (GA) is very slow. To improve this convergent speed is to strengthen its application values. For this, another tool-simultaneous perturbation stochastic approximation (SPSA) for solving optimization problems is introduced. In fact, SPSA is a simple, easily implemented and efficient stochastic approximation algorithm. As a high-speedly convergent and local optimization algorithm, SPSA is mixed with overall searching strategy of GA. Thus, this paper presents a stochastic gradient genetic algorithm (SGGA) for solving multi objective optimization problems (MOP), and specifically designs the new algorithm implementation strategies. Numerical experiments show that SGGA improves the convergent speed of MOP, and finds a large quantity of distributed Pareto-optimal solutions.