具有动态阈值的人工神经网络

Study of the Artificial Neural Network with Dynamic Threshold

在研究神经元阈值作用的基础上 ,提出了神经网络中神经元的动态阈值选择作用和神经元之间抑制作用模型 ,研究表明 ,这种模型能降低应用对象模型的复杂性 ,简化网络运行的中间过程 ,修正二值输出函数的性能 ,使收敛速度和有效性得到改善 .将这种人工神经网络用于求解TSP问题 ,显示系统还具有模拟退火作用的效果 .

Based on the analysis of neuron threshold functions, a model that indicates the selective function of dynamic threshold and the restraint function of ANN neurons is suggested. It is shown in the study that this kind of model can lessen the complexity in objective modeling, simp-lify the process of ANN operation, correct the performance of output function, and then improve the effectiveness and efficiency of the algorithm. The method is used to TSP and it is found that the proposed method has the function of simulated annealing.