Furthermore, the number of hidden nuits, the learning rate, the initial weights and the rule of convergence influence the training of weights. 但是,在运用BP网络时难以找到最佳的权值,此外,隐含层神经元的个数、学习的速率、训练时初始权值以及收敛的方法均影响学习的权值。
The paper sets forward a type of bilevel multi objective decision problem with independent bottom level decision nuits and gives its group partial solution by adopting forward stackelberg superior subordinate decision making mechanism and applying recursive analysis correlating the upper and lower level decision variables. 提出了一类下层决策单元相互独立的两层多目标决策问题,采用正向Stackelberg主从策略的决策机理,利用回归分析关联上、下层的决策变量求解了该类分层决策问题。