隶属DHNN水质评价模型探析  

Analysis of Water Quality Evaluation Model under DHNN

作　　者：路歌 罗义[1] 曹永超 王兆宇 LU Ge;LUO Yi;CAO Yong-chao;WANG Zhao-yu(Hebei Institute of Architecture and Civil Engineering,Zhangjiakou,Hebei 075000)

机构地区：[1]河北建筑工程学院,河北张家口075000

出　　处：《河北建筑工程学院学报》2019年第4期131-135,共5页Journal of Hebei Institute of Architecture and Civil Engineering

基　　金：研究生创新基金(XB201931)。

摘　　要：基于动力学分析设计反馈神经网络Hopfield其离散型的优化算法,通过矩阵分解的方法(MD)得到其正交矩阵,优化算法即可直接计算Hopfield离散型神经网络的权值矩阵,并将样本信息可以较好地存储训练,使测试样本快速收敛到稳定点。相较于传统方法涉及的奇异值分解学习算法不需要分块分步计算,简化了计算步骤和计算量,迭代次数减少,运行速度提高。最后,将训练好的优化算法应用于研究流域生态演变趋势的分析,验证了其应用前景和推广价值。Based on the dynamic analysis,the discrete optimization algorithm of Hopfield neural network is designed in this paper.By matrix decomposition(MD),the orthogonal matrix is obtained,which can directly calculate the weight matrix of Hopfield discrete neural network,and store the sample information for training,so that the test sample can quickly converge to the stable point.Traditional methods need to be calculated in blocks and steps,but the SVD learning algorithm in this paper simplifies the steps and amount of calculation,reduces the number of iterations and improves the running speed.Finally,the trained optimization algorithm is applied in the analysis of the evolution trend of watershed ecology,verifying its application prospect and promotion value.

关 键 词：反馈神经网络 矩阵分解 隶属度 生态演变趋势 

分 类 号：TU990[建筑科学—市政工程]