摘要
针对线性半定规划不可微的问题,将最大熵函数原理应用到半定规划互补问题中,得到扩充的凝聚函数。结合光滑化思想,将半定规划问题的最优条件转化为一个等价的光滑方程组,构造出半定规划的光滑化牛顿法,并证明了该算法的全局收敛性和局部二阶收敛性。
Aiming at nondiffierentiable phenomenon of semidefinite programming（SDP）,maximum entropy function principle was used in SDP complementary problem to get the expanded aggregate function.Combined with the idea of smoothing,the optimal conditions of SDP problem were transformed into an equivalent smooth equation and a smoothing Newton method for SDP was constructed.Finally,both global convergence and local second-order convergence of the algorithm were proved.
出处
《桂林电子科技大学学报》
2012年第5期416-420,共5页
Journal of Guilin University of Electronic Technology
基金
国家自然科学基金(11061011)
广西杰出青年基金(2012GXNSFFA060003)
广西研究生教育创新计划项目(2011105950701M26)
关键词
半定规划
凝聚函数
全局收敛性
二阶收敛性
semidefinite programming
aggregate function
global convergence
second-order convergence
