文章摘要
高畅,孔颖.面向时变复数西尔维斯特方程的有限时间神经网络研究[J].高技术通讯(中文),2022,32(6):587~596
面向时变复数西尔维斯特方程的有限时间神经网络研究
Finite-time neural network study for time-varying complex Sylvester equation
  
DOI:10.3772/j.issn.1002-0470.2022.06.004
中文关键词: 时变复数西尔维斯特方程; 有限时间神经网络(FTNN); 有限值激励函数; 收敛速度; 计算精度
英文关键词: time-varying complex Sylvester equation, finite-time neural network(FTNN), finite-value activation function, convergent speed, calculation precision
基金项目:
作者单位
高畅 (浙江科技学院信息与电子工程学院杭州 310023) 
孔颖 (浙江科技学院信息与电子工程学院杭州 310023) 
摘要点击次数: 973
全文下载次数: 670
中文摘要:
      针对时变复数西尔维斯特方程的实时求解问题,提出了两种有限时间神经网络(FTNN)模型。该方案基于张神经网络(ZNN)在实数域中的动力学方法,设计面向复数域的神经动力学方程。针对动力学方程中非线性激励函数的数值计算问题,应用两种等价的处理方法。第一种方法是处理复数输入的实部与虚部,第二种方法是处理复数输入的模数。通过使用有限值激励函数加快FTNN模型的求解速度,进一步提高了模型的收敛速度和计算精度。实验结果表明,在求解时变复数西尔维斯特方程时,相比于传统的周期神经网络求解法,所提出的网络模型具有更好的收敛性和稳定性。
英文摘要:
      In view of solving the real-time problem of the time varying complex Sylvester equation, two finite-time neural network (FTNN) models are proposed, which is based on the designing rule of Zhang neural network (ZNN) in the real calculation domain. The designing rule of FTNN is aiming at the neural dynamics equation for the complex number domain. Two equivalent methods are applied to the numerical calculation of nonlinear excitation function in dynamic equation. The first method is to deal with the real part and imaginary part of the complex input, and the second method is to deal with the modulus of the complex input. The finite-value activation function is used to accelerate the convergent speed of the FTNN model and obtain high calculation precision. The experimental results show that the proposed network model has better convergence and stability for the solution of time varying complex Sylvester equation compared with the traditional methodology.
查看全文   查看/发表评论  下载PDF阅读器
关闭

分享按钮