Orthogonal eigensubspace estimation using neural networks

We present a neural network (NN) approach for simultaneously estimating all or some of the orthogonal eigenvectors of a symmetric nonindefinite matrix corresponding to its repeated minimum (in magnitude) eigenvalue. This problem has its origin in the constrained minimization framework and has extensive applications in signal processing. We recast this problem into the NN framework by constructing an appropriate energy function which the NN minimizes. The NN is of feedback type with the neurons having sigmoidal activation function. The proposed approach is analyzed to characterize the nature of the minimizers:of the energy function. The main result is that “the matrix W* is a minimizer of the energy function if and only if the columns of W* are the orthogonal eigenvectors with a given norm corresponding to the smallest eigenvalue of the given matrix”. Further, all minimizers are global minimizers. Bounds on the integration time-step that is required to numerically solve the system of differential equations (which define the dynamics of the NN) have also been derived. Results of computer simulations are presented to support our analysis     

An Effective Neural Approach for the Automatic Location of Stator Interturn Faults in Induction Motor

This paper presents a neural approach to detect and locate automatically an interturn short-circuit fault in the stator windings of the induction machine. The fault detection and location are achieved by a feedforward multilayer-perceptron neural network (NN) trained by back propagation. The location process is based on monitoring the three-phase shifts between the line current and the phase voltage of the machine. The required data for training and testing the NN are experimentally generated from a three-phase induction motor with different interturn short-circuit faults. Simulation, as well as experimental, results are presented in this paper to demonstrate the effectiveness of the used method.      

基于BP神经网络的仿真线设计及其FPGA实现

该文提出了一种采用BP神经网络实现仿真线的方法。首先采用遗传算法优化神经网络结构,用离线训练后的BP神经网络逼近传输线的传递函数,然后用STAM算法以较少的存储空间实现BP神经网络的激励函数近似,进而用FPGA和D/A转换器进行硬件实现。文中基于FPGA对长度为10000m,特性阻抗为55的同轴电缆进行了仿真线的硬件实现,实验结果验证了该方法的有效性。该方法可以推广到传递函数未知的传输网络的仿真应用中。     

一种新的特征结构提取方法及其神经网络实现

本文首先建立了特征结构问题的代价函数表示,通过对代价函数求极小可以求得原始数据协方差矩阵的最大特征向量。为了求得其他特征向量,特构造了一个协方差矩阵序列。为实现对代价函数求极小,可把高阶神经网络引入特征结构提取中。这种方法比较直观,它将网络稳定时的输出与所求协方差矩阵的主特征向量的各个分量相对应。理论分析和计算机仿真均验证了这种方法的正确性。     

Neural network-based residual capacity indicator for nickel-metal hydride batteries in electric vehicles

This paper presents a new estimation approach for the battery residual capacity (BRC) indicator in electric vehicles (EVs). The key of this approach is to model the EV battery by using a neural network (NN) with a newly defined output and newly proposed inputs. The inputs are the discharged and regenerative capacity distribution and the temperature. The output is the state of available capacity (SOAC) which represents the BRC. Various SOACs of the nickel-metal hydride (Ni-MH) battery are experimentally investigated under different EV discharge current profiles and temperatures. The corresponding data are recorded to train and verify the proposed NN. The results indicate that the NN can provide an accurate and effective estimation of the BRC. Moreover, this NN can be easily implemented as the BRC indicator or estimator for EVs by using a low-cost microcontroller.     

A Neural Network Method for Traffic Dispersion

１ＩｎｔｒｏｄｕｃｔｉｏｎＴｒａｆｉｃｄｉｓｐｅｒｓｉｏｎ［１］，ｗｈｉｃｈｍｅａｎｓｔｈｅｔｒａｆｉｃｉｓｄｉｓｔｒｉｂｕｔｅｄｏｖｅｒｍｕｌｔｉｐｌｅｐａｔｈｓａｎｄｔｒａｎｓｍｉｔｅｄｉｎｐａｒａｌｅｌ，ｉｓｕｓｅｄｔｏｉｍｐｒｏｖｅｔｈｅｎｅ...     

Absolute stability of analytic neural networks: an approach based on finite trajectory length

A neural network (NN) is said to be convergent (or completely stable) when each trajectory tends to an equilibrium point (a stationary state). A stronger property is that of absolute stability, which means that convergence holds for any choice of the neural network parameters, and any choice of the nonlinear functions, within specified and well characterized sets. In particular, the property of absolute stability requires that the NN be convergent also when, for some parameter values, it possesses nonisolated equilibrium points (e.g., a manifold of equilibria). Such a property, which is really well suited for solving several classes of signal processing tasks in real time, cannot be in general established via the classical LaSalle approach, due to its inherent limitations to study convergence in situations where the NN has nonisolated equilibrium points. A method to address absolute stability is developed, based on proving that the total length of the NN trajectories is finite. A fundamental result on absolute stability is given, under the hypothesis that the NN possesses a Lyapunov function, and the nonlinearities involved (neuron activations, inhibitions, etc.) are modeled by analytic functions. At the core of the proof of finiteness of trajectory length is the use of some basic inequalities for analytic functions due to Lojasiewicz. The result is applicable to a large class of neural networks, which includes the networks proposed by Vidyasagar, the Hopfield neural networks, and the standard cellular NN introduced by Chua and Yang.     

Time Delay Estimation Method Based on Canonical Correlation Analysis

The localization of sources has numerous applications. To find the position of sources, the relative delay between two or more received signals for the direct signal must be determined. The generalized cross-correlation method is the most popular technique; however, an approach based on eigenvalue decomposition (EVD) is another popular one that utilizes the eigenvector of the minimum eigenvalue. The performance of the eigenvalue decomposition (EVD) based method degrades in low SNR and reverberation, because it is difficult to select a single eigenvector for the minimum eigenvalue. In this paper, we propose a new adaptive algorithm based on Canonical Correlation Analysis (CCA) to extend the operation SNR to the lower SNR and reverberation. The proposed algorithm uses an eigenvector that corresponds to the maximum eigenvalue in the generalized eigenvalue equation (GEVD). The estimated eigenvector contains all required information for time delay estimation. We have performed simulations with uncorrelated, correlated noise and reverberation for several SNRs, to show that time delays can be more accurately estimated (especially for low SNR) a CCA based algorithm versus the adaptive EVD algorithm.     

Detection of characteristic waves of sleep EEG by neural network analysis

In psychiatry, the sleep stage is one of the most important evidence for diagnosing mental disease. However, doctors require much labor and skill for diagnosis, so a quantitative and objective method is required for more accurate diagnosis since it depends on the doctor's experience. For this reason, an automatic diagnosis system must be developed. In this paper, we propose a new type of neural network (NN) model referred to as a sleep electroencephalogram (EEG) recognition neural network (SRNN) which enables us to detect several kinds of important characteristic waves in sleep EEG which are necessary for diagnosing sleep stages. Experimental results indicate that the proposed NN model was much more capable than other conventional methods for detecting characteristic waves.     

Design and implementation of industrial neural network controller using backstepping

In this paper, a novel neural network (NN) backstepping controller is modified for application to an industrial motor drive system. A control system structure and NN tuning algorithms are presented that are shown to guarantee stability and performance of the closed-loop system. The NN backstepping controller is implemented on an actual motor drive system using a two-PC control system developed at The University of Texas at Arlington. The implementation results show that the NN backstepping controller is highly effective in controlling the industrial motor drive system. It is also shown that the NN controller gives better results on actual systems than a standard backstepping controller developed assuming full knowledge of the dynamics. Moreover, the NN controller does not require the linear-in-the-parameters assumption or the computation of regression matrices required by standard backstepping.     

Wavelet Neural Network Approach for Testing of Switched-Current Circuits

Combining the time and frequency location and multiple-scale analysis of wavelet transform with the nonlinear mapping and generalizing of neural network, an efficient defect-oriented parametric test method using Wavelet Neural Network (WNN) for switched-current integrated circuits is proposed. Contraposing to the fully compatible digital CMOS technology and current scaling calculation of SI circuits, parameter cohort of switched current elements is used to compute the sensitivity and gain tolerance and is applied for selecting the test models. The selecting of the appropriate wavelet function based on particular switched current fault signal is discussed, and the number of network input and output nodes are determined by the circuit status and dimension of eigenvector which is the energy of wavelet decomposition coefficient. To simplify configuration of the neural network, the sampled data was preprocessed by wavelet transform. Illustrative examples show that the proposed wavelet neural network method for testing of switched current circuits is effective.     

基于MCC的网络流量预测方法

This paper proposes a method for improving the precision of Network Traffic Prediction based on the Maximum Correntropy Criterion (NTPMCC), where the nonlinear characteristics of network traffic are consid-ered. This method utilizes the MCC as a new error evaluation criterion or named the cost function (CF) to train neural networks (NN). MCC is based on a new similarity function (Generalized correlation entropy function, Correntropy), which has as its foundation the Parzen window evaluation and Renyi entropy of error probability density function. At the same time, by combining the MCC with the Mean Square Error (MSE), a mixed evaluation criterion with MCC and MSE is proposed as a cost function of NN training. According to the traffic network characteristics including the nonlinear, non-Gaussian, and mutation, the Elman neural network is trained by MCC and MCC-MSE, and then the trained neural net-work is used as the model for predicting net-work traffic. The simulation results based on the evaluation by Mean Absolute Error (MAE), MSE, and Sum Squared Error (SSE) show that the accuracy of the prediction based on MCC is superior to the results of the Elman neural network with MSE. The overall performance is improved by about 0.0131.     

基于LMBP神经网络的惯性器件故障预报方法研究

利用Levenberg-Marquardt(L-M)算法优化计算BP权值调整量,将L-M算法与传统的BP网络相结合开发出一种快速收敛的LMBP网络,并在此基础上提出了基于LMBP神经网络的时间序列预测方法。最后利用该方法对某惯性器件进行故障预报,通过仿真实验证明了该方法的有效性。     

Assembly yield prediction of plastically encapsulated packages with a large number of manufacturing variables by advanced approximate integration method

An advanced approximate integration scheme called eigenvector dimension reduction (EDR) method is implemented to predict the assembly yield of a plastically encapsulated package. A total of 12 manufacturing input variables are considered during the yield prediction, which is based on the JEDEC reflow flatness requirements. The method calculates the statistical moments of a system response (i.e., warpage) first through dimensional reduction and eigenvector sampling, and a probability density function (PDF) of random responses is constructed subsequently from the statistical moments by a probability estimation method. Only 25 modeling runs are needed to produce an accurate PDF for 12 input variables. The results prove that the EDR provides the numerical efficiency required for the tail-end probability prediction of manufacturing problems with a large number of input variables, while maintaining high accuracy.     

Least-Squares Design of FIR Filters Based on a Compacted Feedback Neural Network

The design of finite-impulse response (FIR) filters can be performed by using neural networks by formulating the objective function to a Lyapunov energy function. Focusing on this goal, the authors present an improved structure of a feedback neural network to implement the least-squares design of FIR filters. In addition to using the closed-form expressions for the synaptic weight matrix and the bias parameter of the Hopfield neural network (HNN), the proposed approach can achieve a notable reduction both in the amount of computation required and hardware complexity compared to the previous neural-based method. Simulation results indicate the effectiveness of the proposed approach     

基于脊波和神经网络的大压缩比遥感图像压缩

为了实现大压缩比的遥感图像压缩,利用神经网络的自组织、并行计算和分布式存储的能力,提出一种基于神经网络的压缩方法.在传统单隐层前向神经网络的基础上,该网络使用一种新的能有效处理直线型和曲线型奇异性的多尺度几何分析工具-脊波,作为隐层神经元的激活函数.它不仅具有神经网络压缩的优点;并且由于脊波良好的时、频和方向局域化特性,能够对遥感图像的边缘和轮廓实现更加有效的表示.仿真结果表明:该方法不仅能实现较高的压缩比,而且具有重建图像质量好、学习快速和鲁棒性强等优点.     

Intracranial Epileptic Seizures Detection Based on an Optimized Neural Network Classifier

Automatic identification of intracranial electroencephalogram (iEEG) signals has become more and more important in the field of medical diagnostics. In this paper, an optimized neural network classifier is proposed based on an improved feature extraction method for the identification of iEEG epileptic seizures. Four kinds of entropy, Sample entropy, Approximate entropy, Shannon entropy, Log energy entropy are extracted from the database as the feature vectors of Neural network (NN) during the identification process. Four kinds of classification tasks, namely Pre-ictal v Post-ictal (CD), Pre-ictal v Epileptic (CE), Post-ictal v Epileptic (DE), Pre-ictal v Post-ictal v Epileptic (CDE), are used to test the effect of our classification method. The experimental results show that our algorithm achieves higher performance in all tasks than previous algorithms. The effect of hidden layer nodes number is investigated by a constructive approach named growth method. We obtain the optimized number ranges of hidden layer nodes for the binary classification problems CD, CE, DE, and the multitask classification problem CDE, respectively.     

基于LSTM神经网络的频谱感知算法

在频谱感知中经典的能量检测算法在低信噪比时检测性能较低且门限难以估计,基于机器学习的感知算法受限于检验统计量的构造会造成接收信号原有结构信息的丢失。针对这些问题,本文提出一种基于LSTM神经网络的频谱感知方法,首先利用接收信号序列作为神经网络的输入特征向量,然后使用LSTM神经网络进行训练得到分类器,最后使用训练好的模型实现频谱感知。该方法无需估计检测门限值,也无需构造特征向量,仿真结果表明,所提算法在采样点和次级用户更少的情况下仍优于对比算法。      

A neural network approach to routing without interference inmultihop radio networks

The issues of routing and scheduling the activation of links in packet radio networks are highly interdependent. The authors consider a form of the problem of routing for the minimization of congestion as a step toward the study of the joint routing-scheduling problem. They formulate this as a combinatorial-optimization problem, and they use Hopfield neural networks (NN) for its solution. The determination of the coefficients in the connection weights is the most critical issue in the design and simulation of Hopfield NN models. They use the method of Lagrange multipliers, which permits these coefficients to vary dynamically along with the evolution of the system state. Extensive software simulation results demonstrate the capability of their approach to determine good sets of routes in large heavily congested networks     

基于模糊神经网络的目标识别

结合模糊推理和神经网络两种方法的优点,从网络的结构、工作过程、学习算法等方面,探讨了一种基于模糊神经网络(FNN)的目标识别方法。通过仿真结果证明,此方法确实可行。