介绍了半定规划的一般模型、最优性条件及求解半定规划问题的原始对偶势下降内点算法。借助两个形象的图形分析了势下降内点算法的迭代轨迹,并对求解半定规划的Filter势下降内点算法进行了研究,提出了Fil-ter的构造方法。在一定的条件下,该算法可避免Maratos效应和势函数海色矩阵不正定等问题的产生。
This paper mainly presents the general model, optimal conditions of the semi-definite programming and its primal-dual potential reduction interior-point algorithm. In particular, two vivid figures are given to help under standing the trace of the iterative points. Moreover, the idea of combining interior-point algorithm with the Filter technique for solving SDP is inspired and the formulation of a practical filter is proposed. Under certain conditions, the new algorithm can avoid the Maratos effects and the non-positive definiteness of the Hessian matrix of potential function.