摘要
鞍点算法在求解大规模线性规划问题时是一种相对较快的算法,但随着维数的增加,解题的复杂性逐渐提高,迭代步骤所需时间变长,精确度变低。分析其产生原因发现步长对收敛性有一定的影响,步长的取值能影响收敛速度和稳定性。在前期研究的基础上,通过选取不同的步长公式,在保证同精度下,研究其收敛速度。研究结果表明:步长的大小影响迭代过程收敛性和收敛速度,当步长1ρ不变时,随着2ρ的增大,迭代过程的收敛速度变快,但稳定性下降。
Saddle point algorithm was a faster method in solving large scale linear programming problem.As the increase of the dimension,the complexity increased and the precision reduced.The reason was that the step length impacted on the convergence,such as the rate and the stability.On the basis of the early study,Change the different step length and research the rate of the convergence. Result:The step length determins the rate and stability.When the step length 1ρ unchanged,with the increase of the step length 2ρ, the speed of the convergence gets faster,but the stability declines.
出处
《科技传播》
2015年第1期174-175,共2页
Public Communication of Science & Technology
关键词
鞍点算法
步长
矩阵特征根
收敛性
Saddle point algorithm
Step length
Matrix characteristic root
Convergence
