利用几何规划的对偶原理,将几何规划问题转化为相应的对偶规划,并利用几何规划及其对偶规划的特点,以及非线性规划共轭梯度算法的研究成果,将2者进行了恰当的结合,构造了无约束正定几何规划的一种有效算法。在算法中采用了一种较好的广义Armijo步长搜索方法,且在较弱的条件下证明了算法的下降性和全局收敛性。
In this paper corres ponding, by using the duality principle of geometric programming, we transformed the geometric programming problem into corres ponding duality programming problem, and using the property of geometric programming and duality programming, as well as the result of conjugate gradient method for non-linear programming, we integrated the two, and proposed a algorithm for unconstrained positive define geometric programming, in that a better generalized Annijo step search method is adopted. We proved the docent property and global convergence of the algorithm under weaker condition.