Markowitz模型的一种神经网络解法  

A Neural Network for Solving Markowitz Model

作　　者：郭文旌[1] 

机构地区：[1]西安电子科技大学经济管理学院,西安710071

出　　处：《兰州铁道学院学报》2002年第3期53-56,共4页Journal of Lanzhou Railway University

基　　金：国家自然科学基金 (199610 0 1)

摘　　要：对不允许卖空情况下的Markowitz模型的求解 ,在金融学里面 ,一直是个很棘手的问题 .提出了一种反馈神经网络算法 .该算法计算步骤简单 ,收敛速度快 ,特别针对大规模问题非常有效 .首先提出了这种算法的神经网络模型 ,给出了它的能量函数 .接着证明了它在Lyapunov意义下的稳定性 .最后证明了一定存在一个收敛序列 {Xk} Rn,使得It has been a very difficult problem to solve Markowitz model with no short sale in finance. In this paper, a feedback neural network algorithm is presented for this problem. The features of this algorithm are simple step, high convergence rate and efficiency. It is especially effective for large_scale problem. Firstly, the neural network model and its energy function is provided. Secondly, the neural network is proven to be Lyapunov stable. Finally, it is proven that there must exist a convergent sequence, so that each limit point of the sequence is the solution of Markowitz model.

关 键 词：MARKOWITZ模型 反馈神经网络 投递组合证券最优化模型 金融学 算法 

分 类 号：F830[经济管理—金融学] TP183[自动化与计算机技术—控制理论与控制工程]