NEURAL MEMORY MODEL AND ITS PIECEWISE-LINEAR NETWORK REALIZATION

This paper studies the neural networks by means of neural functions.The memoryfunction of neural networks is investigated and its mathematical model is given.The model issynthesized by a piecewise-linear resistive network which exhibits many properties of artificialneural network such as parallelism,real-time processing capability,distribution,adaptation.Inaddition,all parameters of the network are expressed analytically by the patterns and featureswhich are memorized in the network.     

Neural network analysis of molecular hyperpolarisabilities

A standard back-propagation neural netwrok is trained to predict the hyperpolarisability β of substituted nitrobenzenses as reported from EFISH experiments. Learning is faster with ¹³C NMR chemical shifts as input than with standard substituent constants and the predictions are somewhat better. The dipole moments μ can be predicted at the same time as β, but training to high precision is then much slower. Developments of this approach may be useful in screening out molecules of high β for synthesis and experimental study.     

基于AI神经网络业务预测的网络容量规划方法研究

本文在无线数据业务需求爆发增长的背景下,基于业务发展趋势提出适用于无线网络容量规划的神经网络预测模型,创新采用数据增量与循环迭代预测方法,提升业务预测精度,并将预测结论应用于无线网络容量规划中,指导网络资源合理配置.     

COLORED NOISE PREDICTION BASED ON NEURAL NETWORK

A method for predicting colored noise by introducing prediction of nonlinear time series is presented. By adopting three kinds of neural networks prediction models, the colored noise prediction is studied through changing the filter bandwidth for stochastic noise and the sampling rate for colored noise. The results show that colored noise can be predicted. The prediction error decreases with the increasing of the sampling rate or the narrowing of the filter bandwidth. If the parameters are selected properly, the prediction precision can meet the requirement of engineering implementation. The results offer a new reference way for increasing the ability for detecting weak signal in signal processing system.     

基于神经网络与马尔可夫组合模型的视频流行度预测算法

为了提升用户体验,降低运营商的成本,将播放最多的视频内容提前放入用户侧缓存是业界的通用做法,如何有效预测视频播放热度已经成为业界热点问题。针对传统预测算法非线性映射能力差、预测精度低及自适应性弱等缺点,提出基于神经网络与马尔可夫组合模型的视频流行度预测算法（Mar-BiLSTM）,该算法通过构建双向长短期记忆（bi-directional long short-term memory,BiLSTM）网络模型可以保留时间序列两个方向的信息依赖;同时在避免引入外部变量导致模型复杂度增加的情况下,利用马尔可夫性质进一步提高了模型的预测精度。实验结果表明,与传统的时间序列和经典的神经网络算法相比,所提算法提升了视频流行度预测的准确性、时效性,并降低了计算量。     

解一般线性规划问题的神经网络

本文提出了一个解一般线性规划问题的神经网络，并在理论上予以了严格证明，对比增广的神经网络，新网络除了同优点外，还具有模型小的特点，最后，给出了网络的电路实现及模拟示例。     

雷达多目标跟踪的神经网络方法

本文讨论了雷达多目标跟踪的神经网络方法。实验结果表明,此法不仅可以解决传统方法带来的组合爆炸问题,从而实现实时的多目标跟踪,而且还可以解决跟踪精度问题,提高跟踪的准确性。     

神经网络用于模式识别的几种主要方法及比较

神经网络的一个主要应用就是模式识别，本文主要讨论了用于模式识别的几种有代表性的神经网络类型及其学习或设计规则，对其优缺点进行了简要分析。     

特征结构的高阶神经网计算

本文首先讨论了特征结构提取问题的代价函数表示，然后以该代价函数为煤介，将高阶关联非线性连续时间神经网络引入到特征结构的提取问题中，这种方法直观、明了，它将所要提取的特征结构与网络稳定时的输出直接对应起来。     

Neural network applications in smart antenna arrays: A review

Techniques employed in the synthesis of antenna arrays vary from complex analytical methods to iterative numerical methods based on optimisation algorithms. The drawback of these techniques is that they usually consider the array factor but not the interaction between array elements and real-time problems. This omission induces an error in the resultant radiation pattern; therefore, the physical relations between the array feeding details and the corresponding radiation patterns are taken into account to improve the accuracy. The behaviour of an antenna array is nonlinear in nature, resulting in an extremely high complexity using this approach, and it is usually disregarded. A neural-network-based solution can avoid complexity by establishing a relation between the desired radiation patterns and feeding details such as voltage and spacing in the real antenna array and can help convert the real array into a smart array. Several neural network applications in smart antenna array synthesis are reviewed in this paper.     

量子神经网络及其应用

量子计算与人工神经网络相结合的量子神经网络(Quantum Neural Networks,QNN)有可能成为未来信息处理的重要手段。分析了人工神经网络向QNN演变的动因及形式、QNN的优势及可能的物理实现方法。着重讨论了几种QNN模型的结构、学习方法及特性,并阐述了QNN在模式识别、纠缠计算、函数近似等方面的初步应用。     

优化计算的神经网络模型

优化计算的动态求解器对于一些需要进行实时分析决策的场合具有很高的实用价值,利用神经网络进行这方面的研究已经引起了广泛的重视。本文介绍了目前国际上具有代表性的几种优化计算网络模型,以及它们的实现方法和存在的问题。     

神经网络优化计算的新方法

本文在Hopfield神经网络优化方法的基础上，根据模拟退火算法逃离局部最优解的原理，提出了一种神经网络优化计算的新方法．通过调整神经网络的连接权，网络的演化不仅可以逃离目标函数的局部最优解，而且可以改善目标函数的局部最优解．实验结果表明，新方法求解最优解所需的计算时间比模拟退火算法少得多．     

约束优化神经网络建模和控制策略研究

根据人工神经网络的基本优化机理，研究了基于Ｌａｇｒａｎｇｅ函数的适合于求解一般约束问题的神经网络建模方法，探讨了神经元非线性度和拉氏乘子等提高网络优化设计效率的控制策略，测试结果证明了提出的网络和控制策略的可行性和有效性。     

改进的神经网络非均匀性校正方法

红外图像的非均匀性是制约红外成像系统成像质量的限制性因素,非均匀性校正是红外成像领域中的核心技术。改进了传统的神经网络非均匀性校正方法,在探测元的期望输出中加入图像均值分量,使该算法对空间相关性较强的低频非均匀性成分也具有较强的校正能力。在以TMS320C62x DSP为核心的硬件平台上,采用中波凝视红外探测器,图像帧频为50 Hz,并在实验室及室外环境分别对该算法进行了测试,非均匀性从最初的5％左右降至2‰以下,验证了改进后的算法具有校正精度高、实用性强等优点,能够满足实际应用的需要。该算法也可应用于红外成像的应用领域。     

天线阵CDMA系统中基于神经网络的盲空时信道估计

提出了天线阵ＣＤＭＡ系统中盲空时信道估计的约束优化神经网络模型,对其全局收敛性进行了分析,并对其性能进行了数值模拟。     

用于光流计算的一种神经网络算法

本文给出了一种利用神经网络计算光流场的新算法。整个计算过程分为三个阶段:神经网络模型参数的估计,轮廓边界垂直速度分量的动态测量以及光流场的计算。通过网络能量函数和运动的约束误差函数的比较对网络参数进行估计。用一个动态算法迭代运行非线性光流场计算方法以使神经网络能量函数达到最小,同时也对垂直速度分量进行动态估计。由模拟试验结果讨论了影响神经网络收敛性能的若干因素。     

神经网络在手写体数字识别中的应用

本文采用前向、多层神经网络,BP学习算法对40个人的手写体数字进行了识别。识别过程分为四步:首先,用HP扫描仪把写在纸上的数字变成二值图像,接着对它进行分割,规整等预处理,变换成32×32点阵。然后提取特征,把点阵图像变成特征描述。最后,进行训练和识别。在拒识率为25％条件下,得到误识率为0.4％的识别结果,文中还分析和讨论了在实验中遇到的一些问题。     

Neural network based adaptable control method for linearization of high power amplifiers

The proposed adaptable control method for linearization of high power amplifiers is powered by the neural networks technique that supports analogue polynomial type of predistorters, which are widely utilized in commercial power amplifiers for wireless communication purposes. This paper presents an algorithm to determine the coefficients by use of artificial neural networks and its generalization feature that helps to map the power amplifier response with optimal coefficients of the polynomial, which in a proper way pre-distorts an input signal of the amplifier. The concept of the predistortion has been introduced. Furthermore, the overall step-by-step initialization and functionality of the control process has also been described. The method has been tested successfully in a real power amplifier equipped with an analogue predistorting circuits. Presented measurement results imply that this approach is robust and well suited for such category of the power amplifier design, in which the artificial neural networks play substantial role.     

Hopfield型联想记忆神经网络一种新的分析方法

本文通过定义一种新的能量函数，分析了Ｈｏｐｆｉｅｌｄ型神经网络的渐近稳定性与联想记忆问题，得到了四组保证网络平衡点是渐近稳定平衡点的充分条件，应用这些条件，便可设计联想记忆神经网络，文中给出了应用这些结论设计联想记忆神经网络的实验结果及分析。