一种改进的RBF全局优化方法

径向基函数（Radial Basis Functions）由于具有良好的近似效果和运算简单的特点，被应用于全局优化中，成为解决黑箱函数全局优化问题的有效方法。然而现有的基于RBF的全局优化算法存在迭代过程中RBF模型重构效率低下，以及采样方法不合理导致函数估值次数过多等问题。在此提出几个改进思路：采用基于矩阵分块的增量RBF方法以减少模型重构时间提高效率；采用增量LHD采样方法以确保具有更好的空间填充性；采用算法重启策略以降低估值次数。通过实验验证改进方法的优势。

Improved global optimization method using radial basis functions.

RBF(Radial Basis Functions)due to good effect on approximating and the characteristics of simple operation, is applied to global optimization in solving black-box functions as an effective method. However, the existing global optimization algorithms based on RBF have several shortcomings, such as the inefficiency on RBF surrogate reconstruction in iteration process, large number of evaluations caused by unreasonable sampling and so on. Several improvement ideas are proposed: a block matrix based method is used to reduce the time of the RBF surrogate reconstruction; an incremental LHD sampling is used to get a better space filling; a restart strategy is used to reduce the number of evaluations. The advantages of the improved algorithm are proved with experiments.