一种求解全局优化问题的修正相对熵算法及其收敛性

A Modified Cross-Entropy Method for Solving Continuous Optimization Problems and Its Asymptotical Convergence

相对熵算法是一种应用随机模拟技术求解组合与连续全局优化问题的高效率方法.本文给出一个修正的求解全局优化问题的相对熵算法(MCE),并在假设所求解的问题仅有一个全局最优点的条件下,给出了修正算法的渐近收敛性.数值结果显示,MCE算法至少和CE算法具有同样的有效性.

The Cross-Entropy Method （CE） is a method with great efficiency, using random simulation technique, for combinatorial and continuous global optimization. In this paper, a modification of the cross-entropy method （MCE） for finding the solution of continuous global optimization problem is given. Furthermore, asymptotical convergence of the MCE method is proved, when objective function of the problem may have several local optimizers but only one of them is the global optimal solution over its feasible region.