前馈神经网络的一种有效学习算法

本文提出了基于混合ＧＮ－ＢＦＧＳ法进行前馈神经网络学习的新算法，该算法结合ＧＮ法与ＢＦＧＳ法的特点，既利用了问题本身的特殊结构，又能取得超线性甚至二次渐近收敛率。与ＢＰ算法相比，这种算法可取得更快和更可靠的学习特性，在学习过程中利用该方法能够区分非零残量和零残量问题的特点，提出了自动调整隐单元数的方法，从而可以保证网络学习与归纳能力，示例系统的结果表明了所提方法的有效性。     

一种改进的量子神经网络训练算法

针对训练多层激励函数量子神经网络(MAF-QNN)时权值与量子间隔的目标函数存在冲突,导致收敛速度和网络性能下降的问题,该文提出一种改进的量子神经网络的训练算法。通过设计输出均方误差和这一目标函数对权值和量子间隔进行统一训练,同时引入Levenberg-Marquardt(LM)算法降低目标函数陷入局部极小值的概率,实现了对量子神经网络的高效训练。实验结果表明,该文提出的训练算法有效减少了迭代次数,显著提高了网络收敛精度,可应用于数据分类、函数逼近等场合,扩展了多层激励函数量子神经网络的应用领域。     

Exponential Stability for Delayed Cellular Neural Networks

The exponential stability of the delayed cellular neural networks （DCNN＇s） is investigated. By dividing the network state variables into some parts according to the characters of the neural networks, some new sufficient conditions of exponential stability are derived via constructing a Liapunov function. It is shown that the conditions differ from previous ones. The new conditions, which are associated with some initial value, are represented by some blocks of the interconnection matrix.     

Stability for Cellular Neural Networks with Delay

The cellular neural networks with delay (DCNN‘s) are investigated, and some new sufficient conditions on asymptotical stability of DCNN‘s are derived by constructing the Liapunov functional and utilizing M-matrix and the ω-limit set. It is shown that the new conditions are not related to the delayed parameter.     

分形细胞神经网络用于联想记忆

本文提出了分形细胞神经网络，并成功地应用于联想记忆，从模拟结果看，分形细胞神经网络的联想记忆能力好于Ｂａｒａｍ提出的分形神经网络。     

Computing with Front Propagation: Active Contour And Skeleton Models In Continuous-Time CNN

In this paper, a linear CNN template class is studied with a symmetric feedback matrix capable of generating trigger-waves, a special type of binary traveling-wave. The qualitative properties of these waves are examined and some simple control strategies are derived based on modifying the bias and feedback terms in a CNN template. It is shown that a properly controlled wave-front can be efficiently used in segmentation, shape and structure detection/recovery tasks. Shape is represented by the contour of an evolving front. An algorithmic framework is discussed that incorporates bias controlled trigger-waves in tracking the active contour of the objects during rigid and non-rigid motion. The object skeleton (structure) is obtained as a composition of stable annihilation lines formed during the collision of trigger wave-fronts. The shortest path problem in a binary labyrinth is also formulated as a special type of skeletonization task and solved by combined trigger-wave based techniques.     

Cellular Neural Networks With Transient Chaos

A new model of cellular neural networks (CNNs) with transient chaos is proposed by adding negative self-feedbacks into CNNs after transforming the dynamic equation to discrete time via Euler's method. The simulation on the single neuron model shows stable fix points, bifurcation and chaos. Hence, this new CNN model has richer and more flexible dynamics, and therefore may possess better capabilities of solving various problems, compared to the conventional CNN with only stable dynamics     

An Algorithm for Optimal Resource Allocation in Cellular Networks with Elastic Traffic

In this letter we propose a power allocation iteration which optimizes the weighted aggregate performance of a single-hop network. We show that the proposed iteration is a competitive alternative to conventional gradient iterations in terms of convergence and computational effort.     

On Stability of Cellular Neural Networks

The main results about stability of cellular neural networks (CNNs) are reviewed. Some of them are extended and reformulated, with the purpose of providing to the CNN designer simple criteria for checking the stability properties. A particular emphasis is given to the conditions for the stability of CNNs described by space-invariant templates.     

广义的时滞细胞神经网络的动态分析

本文研究了广义细胞神经网络的动态行为,首先利用度理论建立了客中广义时滞细胞神经网络的平衡点存在唯一物充要条件,其次给出这种广义时滞神经网络全局指数稳定的充要条件。     

A Dedicated Multi-Chip Programmable System for Cellular Neural Networks

Cellular Neural Networks (CNN's) represent a remarkable improvement in the hardware implementation of Artificial Neural Networks (ANN's). In fact, their regular structure and their local connectivity feature contribute to render this class of neural networks especially appealing for VLSI implementations. CNNs are widely applied in several fields, including image processing and pattern recognition. In this research, the authors already presented two fully digitally programmable CNN chips with 3×3 (3×3DPCNN chip) and 6×6 cells (6×6DPCNN chip) respectively. In this paper, a system with twenty of the latter chips will be presented. The main features of this electronic system consist of the full digital programmability of the templates, the digital input/output for logic operations, the analog outputs for dynamic analysis and the implementation of space-variant as well as space-invariant CNNs.     

活动轮廓模型综述

基于活动轮廓模型的目标分割、物体跟踪方法是近十几年来图像和视频领域研究的热点,它可以将待处理问题的先验知识与各种图像处理算法有效地融合在一起,比以往的计算机视觉理论有更强的实用性.本文结合图像分割方法从指导思想和所用的数学方法两方面对活动轮廓模型特别是几何活动轮廓模型中基于水平集方法的C-V方法做了一定综述.     

D-CAT: An Efficient Algorithm for Distributed Channel Allocation in Cellular Mobile Networks

We propose a distributed channel allocation algorithm based on a threshold scheme, called D-CAT, for cellular mobile networks. The algorithm employs two thresholds: (i) a heavy threshold for determining whether a cell is heavy, or overloaded, and for triggering the channel allocation algorithm; and (ii) a target threshold for indicating the target number of free channels that a heavy cell intends to acquire. Based on the two-threshold scheme, the D-CAT algorithm can determine the optimal number of free channels as well as the cell(s) from where a heavy cell should import channels in order to satisfy the required channel demand. Simulation experiments and analyses show that the proposed algorithm incurs lower overhead for channel allocation and is more efficient in terms of channel utilization than other distributed channel allocation algorithms. It also outperforms other centralized and distributed algorithms in terms of call blocking probability.     

基于神经网络的红外焦平面光学非均匀性校正改进算法

基于场景的非均匀校正依然是红外领域的一个研究热门.神经网络算法是一种较为典型的场景校正算法.本文主要针对神经网络算法本身不能校正光学引入的非均匀性问题,提出了新的改进算法,通过对神经网络输入层的预处理,消除图像的低频噪声,此外,为了消除预处理对图像对比度的影响,本文增加了神经网络的层数,使用双层神经网络对算法进行更新,从而消除了图像对比度下降的现象.实验结果表明,改进的神经网络算法能够有效的改善图像质量,消除图像中光学引入的非均匀性.     

基于交替投影神经网络的带限信号外推算法

本文通过对频带受限数字信号的离散傅立叶变换特性的研究,引进了交替投影神经网络,并将其应用范围从实数域拓广到复数域,且给出了在复数域仍然成立的若干结论.运用这些结论,在对网络噪声抑制、网络收敛速度及待外推信号因截断而造成频谱严重外泄等问题的分析与讨论的基础上,提出了一种基于交替投影神经网络的外推算法.仿真实验表明该方法是行之有效的.另外,该算法对频谱外推同样适用;由于它采用全互连神经网络结构,易于并行计算和VLSI实现,从而可满足军事上实时处理的需要.     

神经网络自适应有源噪声对消器

本文提出一种新颖的非线性自适应滤波器-递归神经网络(TNN)自适应滤波器．讨论RNN滤波器的构成及算法,实现了RNN自适应有源噪声对消器．RNN自适应有源噪声对消器性能优于由线性滤波器组成的线性对消器,能适应各种噪声环境．仿真结果充分证明了这一点．     

带阶梯输出函数的细胞神经网络

本文给出了带阶梯输出函数的细胞神经网络的稳定性定理。利用阶梯输出函数的各“台阶”记忆不同灰度，实现了灰度模式的ＣＮＮ联想记忆。     

具有时滞的细胞神经网络的稳定性

本文研究具有时滞的细胞神经网络的稳定性问题，利用构造李雅普诺夫泛函、常数变易法及不等式分析技巧，给出了时滞细胞神经网络全局渐近稳定性的充分条件。     

一种基于主动轮廓模型的自适应模板更新算法

在目标相关跟踪过程中,由于目标的姿态、大小发生变化,所以合理地更新模板极为重要.而已有的模板更新方法都不能适应目标的姿态和大小变化.提出了一种模板更新的新方法,该方法以颜色直方图的Bhattacharyya系数为基础,来进行模板更新时机的判断.并利用主动轮廓模型算法得到目标的边缘,自适应调整模板的大小和内容,从而实现对目标在发生姿态和大小等变化下的可靠跟踪.     

随机细胞神经网络的指数稳定性

本文研究了随机时滞细胞神经网络的指数稳定性,建立了这种细胞神经网络均方指数稳定与几乎必然指数稳定的充分条件,其条件分为时滞无关与时滞相关两种.