一类二层决策问题的性质分析和基于Frank-Walfe方法的神经网络算法

A NEW ALGORITHM BASED ON THE FRANK-WOLFE METHOD AND NEURAL NETWORK FOR A CLASS OF BILEVEL DECISSION MAKING PROBLEMS

提出一类二层决策问题的模型．在某些凸性假设下，研究它的一些性质，提出了一种基于Ｆｒａｎｋ－Ｗａｌｆｅ和人工神经网络方法的算法．并以算例说明该算法可行性．

In this paper a model for a class of bilevel decision-making problems is presented in the form of a static Stackelberg game. Under certain convexity assumptions,some properties of the model is studied. According to these properties, this paper presents a new algorithm based on the Frank-Wolf e method and neural network method. An example demonstrates the feasibility of the proposed algorithm.