本文首先利用对偶理论将正定式几何规划转化为带有非负约束和线性等式约束下的一般非线性规划问题:其次将广义梯度投影算法与内点算法相结合构造出了广义梯度投影内点算法来求解这个非线性规划问题;最后进一步证明了这种算法的收敛性质。此算法不需要计算与跟踪主动约束集,减少了计算量。
In this paper, we firstly change the postive-type geometric-programming into the usual nonliner programming by making use of the dual theory; by integrating the general gradient-projection algorithm and the interior-point algorithm, a new algorithm is derived to solve the nonliner program- ming; last, we prove the convergence of the designed algorithm. The algorithm does not need to know and track the active constraint set, which reduces its complexity.