摘要
基于问题的结构特点,提出了求解一类凸二次极大极小问题的一个新的神经网络.定义了恰当的能量函数,严格证明了该网络是Lyapunov稳定的,并且大范围渐近收敛于原问题的一个精确解.此外,新模型在适当的条件下是指数稳定的.由于新模型的规模与原问题相同,并且参数易于选择,因此其结构简单,更适合于硬件实现.数值试验表明新模型不仅可行,而且有效.
Based on its inherent properties,a new neural-network model for solving convex quadratic minimax problems is proposed in this paper.The new model is proved to be stable in the sense of Lyapunov, and convergent to an exact saddle point by defining an energy function.Furthermore,the exponential stability of the new model is also shown under certain conditions.Since the size of the new model is same as that of the original problem,and its network's parameter can be easily chosen,its structure is simpler,and it is more suitable to be implemented by in simple hardware.The validity and transient behavior of the proposing neural network are demonstrated by some simulation results.
出处
《纺织高校基础科学学报》
CAS
2005年第2期143-147,154,共6页
Basic Sciences Journal of Textile Universities
基金
陕西省师范大学重点科研项目(995091)
关键词
极大极小问题
收敛性
稳定性
神经网络
convergence
stabiltiy
minimax problem
saddle point
neural network
