线性规划单纯形法的改进—避免无效回溯和循环.docxVIP

线性规划单纯形法的改进—避免无效回溯和循环余四清（北京市丰台科学城富丰路6号北京金自天正公司轧钢部 100071 e-mail: yusiqing@126.com）摘要对于用线性规划单纯形法解决问题时所遇到的一类无效回溯和循环现象，本文提出了简单而有效的解决方法，即尽量避免以等于零的变量入基，改进了单纯形法选择入基变量的方法。这种方法对减少因入基变量为零而产生的回溯现象很有效，减少了计算量；它打破了Klee和Minty给出的实例叠代次数为变量个数的指数函数的结论，具有极其重要的理论意义。关键词线性规划，单纯形法，多项式时间算法，回溯现象，循环现象分类号O211.1Improvement of Simplex Algorithm in Linear Programming --Avoid Ineffective Recalling and CirclingYU Siqing(Steel-Rolling Department, Beijing AriTime Intelligent Control Co., Ltd., 6Fufeng Road, Fengtai Science City,Beijing 100071)Abstract For ineffective recalling and circling phenomenain linear programming using simplex algorithm, a simple and effective solving method is put forward, that is avoiding to use variable whose value is zero as variable which will enter the basic matrix. It improves the method of selecting variable which will enter the basic matrix in simplex algorithm. The method is effective in reducing the recalling phenomena because of the zero-value variable which will enter the basic matrix. It reduces the calculating amount. It breaks the result of an example which were given by Klee and Minty that the substitution times is an exponential function of the number of variables and has very important theoretical significance.Key Words linear programming, simplex algorithm, polynomial-time algorithm, recalling phenomena, circling phenomena引言对于线性规划问题，G. B. Dantzig1947年提出的单纯形法是非常有效的，但它不是多项式时间算法。所谓多项式时间算法，就是算法计算时间不超过将问题的数据输入计算机上所需二进制代码长度的某个多项式所确定的数值的算法。原因是Klee和Minty[1]给出实例：min f= –xnst. x1 - r1 =1/4 x1 + r1 =1 xj –1/4 xj - rj =0，j=2,3,…,n xj +1/4 xj + sj =0，j=2,3,…,n xj , rj , sj ≥0，j=1,2,…,n文献[1]证明，用单纯形法解之，叠代次数为2n-1。另外，Beale[2]给出的例子：min f=-3/4 x4 + 20x5 – 1/2x6 + 6x7st. x1 +1/4 x4 - 8 x5 - x6 +9 x7 = 0 x2 +1/2 x4 -12 x5 –1/2x6 +3 x7 = 0 x3 +x6 = 1xj≥0，j=1,2,…,7用单纯形法解之，会产生无限循环现象，得不出结论，需要用所谓摄动法解才能得出结论。本人认为，产生以上结果的原因是叠代过程中产生了回溯现象。在每次叠代过程中，要选择一个入基变量和一个离基变量，以形成新的基。如果某个变量入基后又离基，或离基后又入基，这就是回溯现象。试看Beale的例子单