论文部分内容阅读
针对离散Hopfield神经网络(DHNN)的权值设计问题,提出一种改进型学习算法,并在DHNN动力学分析的基础上设计该学习算法.利用矩阵分解的方法(MD)得到正交矩阵,并采用得到的正交矩阵直接计算DHNN的权值矩阵.通过该学习算法得到的权值矩阵,可以很好地存储训练样本的信息,使测试样本收敛到稳定点.该学习算法不需要进行分块计算,减少了计算步骤和计算量,降低了网络的迭代次数,从而提高了网络运行速度.最后,将该学习算法应用于水质评价,验证了其有效性和可行性.
In order to solve the problem of weighting discrete Hopfield neural network (DHNN), this paper proposes an improved learning algorithm and designs the learning algorithm based on DHNN dynamic analysis. The orthogonal matrix is obtained by matrix decomposition (MD) The weight matrix of DHNN can be directly calculated by using the orthogonal matrix obtained, and the weight matrix obtained by the learning algorithm can well store the information of the training samples so that the test samples converge to the stable points. The learning algorithm does not need to be subdivided Which reduces computation steps and computational cost, reduces network iteration times and improves network running speed.Finally, this learning algorithm is applied to water quality evaluation to verify its effectiveness and feasibility.