一种全局优化的随机水平值逼近算法

Stochastic Level-Value Approximation for Global Optimization

对一类在闭箱上处处有定义的单峰目标函数的全局优化问题,提出一种随机水平值逼近算法,证明了算法的渐近收敛性.数值结果验证了算法的有效性.

We propose a stochastic level-value approximating method （SLVAM） for global optimization problem that is defined in an entire box. We discuss and establish a sufficient condition for the global optimality in probability measure, and prove that the SLVAM algorithm is convergent in probability. Numerical results show its effectiveness.