为求解线性二阶锥规划,介绍了一种修正FR共轭梯度法。给出线性二阶锥规划问题的KKT条件,利用F-B光滑函数将互补性条件光滑化,将KKT条件转化成一个与之等价的光滑非线性方程组,给出一个价值函数,将光滑非线性方程组转化为无约束优化问题,利用共轭梯度法求解无约束优化问题,得到原问题的最优解。证明该算法的全局收敛性。
A modified FR conjugate gradient method is presented for solving linear second-order cone programming. Firstly, the Kar ush-Kuhn-Tucker(KKT) optimality condition of linear second-order cone programming is given. The complementarity in the KKT optimality condition can be smoothed by the F-B smoothing function. The KKT optimality condition can be transformed into a equivalent smoothing nonlinear equation. It was transformed into an unconstrained optimization problem by a value function. The unconstrained optimization problem can be solved by conjugate gradient method, then the optimal solution of primal problem can be got. Finally, the analysis of the global convergence properties of the algorithm is proved.