针对传统算法无法获得凸可行问题多个可行解的困难,结合差分进化算法,提出了求解凸可行问题的一种光滑型方法.首先利用投影函数将凸可行问题转换为非光滑方程组,然后用光滑函数对其进行光滑化,把凸可行问题转换为无约束优化问题,利用差分进化算法进行求解.该算法对目标函数的解析性质没有要求,易于实现,数值结果表明了该方法的有效性.
For the difficulty that the classical algorithms are unable to obtain many solutions to convex feasibility problem, the smoothing algorithm, differential evolution algorithm, for convex feasibility problem is proposed in this paper. First, reformulated convex feasibility problem as a sys tem of nonsmooth equations via a projection function, given a smoothing approximation to the nons mooth equations by smoothing function, furthermore, convex feasibility problem could be trans formed into unconstrained optimization, then using the differential evolution algorithm to solve this problem. The objective function of the algorithm does not require analytical nature and easy to ob tain, numerical results show that the method is effective in solving convex fe.asibillty problem.