对凸二次整数极小化问题提出了一种随机水平值逼近算法,该算法应用了重点取样技术,并利用极小化相对熵的思想来更新取样密度.对算法的渐近收敛性进行了证明,给出了数值实验的结果.
A stochastic level value approximating method for quadratic integer convex minimizing problem was proposed. This method applies the importance sampling technique, and uses the main idea of the cross-entropy method to update the sample density functions. The asymptotic convergence of this algorithm was also proved, and some numerical results to illuminate its efficiency was reported.