具有参数摄动的时滞Hopfield神经网络的鲁棒稳定性

 研究一类具有参数摄动的时滞Hopfield神经网络模型的鲁棒稳定性.应用Lyapunov泛函法,给出了平衡点渐近稳定的充分条件.利用矩阵范数的性质及线性矩阵不等式(LMI)理论,又得到了两个便于计算和验证的推论.提供了一种估计网络渐近稳定平衡点吸引域的方法,并详尽地分析了吸引域对神经网络实现联想记忆的影响.数值例子进一步证明了结论的有效性.     

具有无穷时滞的细胞神经网络的稳定性分析

本文研究了具有无穷时滞的细胞神经网络的全局吸引性问题.利用常数变易法和不等式分析技巧,给出了无穷时滞的细胞神经网络无平衡点时,网络系统有吸引紧集的充分条件,同时也给出了无穷时滞的细胞神经网络有平衡点时,网络系统的平衡点全局渐近稳定的充分条件.其结果推广了文 的相应结果.     

随机变时滞神经网络的全局渐近稳定性

研究了一类具有变时滞随机神经网络模型平衡点的全局渐近稳定性问题,通过构造李亚普诺夫函数并利用线性矩阵不等式理论,得出了随机变时滞神经网络的全局渐近稳定性的充分条件。     

对称神经元网络：渐近稳定性和学习

本文在分析对称神经元网络的渐近稳定性基础上,提出了一种非约束优化学习算法,保证训练样本成为稳定吸引子,具有一定大小的吸引域.理论上,我们证明了算法的收敛性以及形成的吸引域下界.计算机实验结果充分说明了学习算法的优越性.     

基于神经网络吸引子纠错译码的初步研究

介绍了利用神经网络吸引子译码的新思想和一种吸引域的分析方法,并给出了一个吸引子译码的具体例子.通过计算机模拟说明了吸引子译码的可行性,同时指出了吸引子译码的局限性.     

扩展交替投影神经网络 ——具备联想记忆功能的充分必要条件

对交替投影神经网络(APNN)的连接权矩阵进行修改,将其应用范围从实数域拓展到复数域,从而得到一种新的神经网络——扩展交替投影神经网络(Extended Alternating Projection Neural Networks).对EAPNN网络进行深入研究后,给出了网络稳态值的通用数学表达式,并从表达式中推出了网络具备联想记忆功能的充分必要条件.最后设计仿真实验对文中的理论分析结果进行了验证.     

中立型BAM神经网络的鲁棒稳定性分析

本文主要研究不确定中立型BAM神经网络的鲁棒渐近稳定性问题, 不确定参数具有较范数有界更一般的线性分式形式, 考虑了中立时滞与状态时滞不相等的情况, 激励函数只要求满足有界和全局李普希兹条件,通过构造一个新的Lyapunov泛函, 利用Lyapunov-Krasovskii稳定性理论和一些不等式技术, 得到了具有较小约束的时滞中立型BAM神经网络的鲁棒渐近稳定性条件, 这个充分条件以线性矩阵不等式的形式给出, 容易验证.最后, 通过数值实例验证了所提算法的正确性和保守性.     

基于一类T-模的模糊Hopfield网络的稳定性分析

本文建立了基于一类T-模的一种动态模糊神经网络-(∨,T)Hopfield网.文章首先证明了系统本身的稳定性及系统平衡态的Lyapunov稳定性,然后建立了吸引子的一个非平凡吸引域,使系统具有良好的容错性.最后,通过例子验证了得到的结论.     

具有变时滞的静态神经网络新的稳定性条件

考虑一类具有变时滞的静态神经网络的渐近稳定性问题,基于Lyapunov稳定性理论,时滞分解的思想,并利用时滞导数的上下界,得到了线性矩阵不等式表示的新的渐近稳定性条件,最后,两个数值例子表明所得结果较一些现存结果具有更小的保守性。     

结合ICA和复数神经网络的二麦阵列盲源分离技术

为了降低语音信号盲源分离算法的延时,提高其准确性和稳定性,本文结合传统盲源分离技术和深度神经网络的优势,提出了一种基于ICA独立分量分析和复数神经网络的二麦阵列盲源分离技术。本文将复数递归神经网络和独立分量分析方法有机融合,提出一种基于时频域的双通道复数神经网络,同时解决了独立分量分析中的排列问题。所提方法利输入混合信号利用复数域神经网络计算初始化分离矩阵,神经网络输出采用复数域形式,利用复数学习标签估计复数矩阵,然后采用独立分量分析方法获得目标分离矩阵。实验数据表明,所提方法相较于其它独立分量分析方法提高了盲源分离的实时性和准确性。      

高阶Hopfield型神经网络的吸引域和指数收敛速度的估计

利用 Lyapunov方法,对高阶连续 Hopfield型神经网络平衡点的局部指数稳定性进行研究,得到了该网络平衡点的吸引域及其中每点趋向于平衡点的指数收敛速度的估计结果。     

Power system transient stability analysis: Formulation as nearly Hamiltonian systems

In this paper we formulate power systems as nonlinear nearly Hamiltonian systems. Using the invariance principle for ordinary differential equations, necessary and sufficient conditions for asymptotic stability are established and a new method of estimating the domain of attraction of the stable equilibrium point is developed. The present results constitute a novel approach to stability analysis and involve the following three steps:

1. Given a system with dissipation, the stability of its equilibrium is ascertained by determining the stability of the associated conservative system.
2. Attractivity of the stable equilibrium of the entire system (with dissipation) is determined from the system topology.
3. An estimate of the domain of attraction of the asymptotically stable equilibrium is obtained by making use of results obtained in (a) and (b).

The stability criterion developed in this paper sheds new light on the mechanism of instability in power systems and it provides analytical verification to the concept of the potential-energy boundary surface (PEBS). The PEBS is a hypersurface which makes up a part of the boundary of the domain of attraction of the stable equilibrium in a power system. The existence and properties of the PEBS have thus far been deduced primarily via simulations and heuristic methods.     

Stabilization of Bilinear Systems Via Linear State-Feedback Control

This paper deals with the problem of stabilizing a bilinear system via linear state-feedback control. The proposed procedures enable us to compute a static state-feedback controller such that the zero-equilibrium point of the closed-loop system is asymptotically stable; moreover, it ensure that an assigned polytopic region is enclosed into the domain of attraction of the equilibrium point. The controller design requires the solution of a convex optimization problem involving linear matrix inequalities. The applicability of the technique is illustrated through an example, dealing with the design of a controller for a Ćuk dc–dc converter.      

A Frequency Domain Method for Analyzing Period Doubling Bifurcations in Discrete-Time Systems

A frequency domain methodology for the approximation of period doubling bifurcations in discrete-time nonlinear systems is presented. The computation of a stability index, which characterizes the dynamical behavior of the emerging period-two orbit is also derived. The technique is applied to estimate the domain of attraction of the fixed point in an adaptive control system and to improve the dynamical behavior of a buck converter.     

Global Exponential Stability Analysis of a Class of Dynamical Neural Networks

The problem of the global exponential stability of a class of Hopfield neural networks is considered. Based on nonnegative matrix theory, a sufficient condition for the existence, uniqueness and global exponential stability of the equilibrium point is presented. And the upper bound for the degree of exponential stability is given. Moreover, a simulation is given to show the effectiveness of the result.     

Global Exponential Stability Analysis of a Class of Dynamical Neural Networks

The problem of the global exponential stability of a class of Hopfield neural networks is considered. Based on nonnegative matrix theory, a sufficient condition for the existence, uniqueness and global exponential stability of the equilibrium point is presented. And the upper bound for the degree of exponential stability is given. Moreover, a simulation is given to show the effectiveness of the result.     

Global asymptotic stability of cellular neural networks with unequal delays: LMI approach

A criterion for the global asymptotic stability and uniqueness of the equilibrium point of cellular neural networks with unequal delays is presented. The criterion is computationally efficient, since it is in the form of linear matrix inequality (LMI).     

多时滞Cohen-Grossberg神经网络鲁棒稳定性的新判据

分析了一类具有多时滞及参数摄动的Cohen-Grossberg神经网络的全局鲁棒稳定性.通过构造适当的Lyapunov泛函,以线性矩阵不等式形式给出了平衡点全局鲁棒稳定的判据.此外,所有结论的建立都不需要假定互连矩阵的对称性及激活函数的可微性和单调性.仿真结果进一步证明了结论的有效性.     

Improved Global Robust Stability Criteria for Delayed Neural Networks

New criteria for the uniqueness and global robust stability of the equilibrium point of the interval Hopfield-type delayed neural networks are presented in the form of linear matrix inequality. An example is given to show the effectiveness of the obtained results.