共轭方向直接法及其收敛性

本文给出一种新的直接法，构造一组特定的共轭方向，使其只需沿一个方向进行一维搜索，比通常的共轭方向法节省了许多一维搜索的工作量，从数值计算上看，迭代次数较Ｐｏｗｅｌｌ法有明显减少。

A Conjugate Direct Search Algorithm and Its Convergence

In the paper, a new direct search algorithm for unconstrained optimization is given. We design a series of the special conjugate directions, so that the algorithm linear searchs along only one direction in that. The algorithm is less than general conjugate direction algorithm in the number of one -dimension search. The convergence of the algorithm is shown. Many numerical examples show that the algorithm is efficient and the number of iterations is evidently decrease.