基于视中枢神经机制的层次网络计算模型

在视皮层区中,有许多非常规整的柱形功能结构,它们形成的局域网络具有抽取视图像中最基本的特征的计算能力,相邻视神细胞则抑制机制和神经细胞的感受野为实现这样的并行计算能力提供了保证,并且以层的这种等级组构为许多心理现像提供了生理解释,这不仅对模式识别、计算机视觉有重要的价值,而且对人工智能系统的知识获取和知识表示都具有非常重要的意义,文中通过构造一个金字塔状的神经网络层次模型,来对模拟视网膜的输入点阵     

Evolutionary learning based sustainable strain sensing model for structural health monitoring of high-rise buildings

Strain sensor network-based structural health monitoring systems have been used to assess the safety of high-rise buildings. In consideration of life cycle of high-rise buildings, long-term measurement by sensors should be required. However, because of unpredictable problems such as the lack of durability of sensors and data loggers, disruption in communication, and loss of data, long-term strain measurement of major structural members is currently infeasible. For sustainable safety assessment of high-rise buildings, this paper presents a sustainable strain-sensing model that employs an artificial neural network (ANN) to estimate the strain responses of columns depending on the wind-induced behavior of high-rise buildings. The ANN model used in the paper is based on evolutionary learning consists of training in radial basis function neural network (RBFN) and evolving in genetic algorithm. In this evolutionary RBFN (ERBFN). Weights between layers are trained and variables of Gaussian function in the RBFN are evolved to estimate strain responses of the column of the high-rise building structure. A wind tunnel test was performed to produce wind data and strains in column members in a high-rise building model. In the wind tunnel test, a specimen consisting of a core, perimeter columns, and outriggers is used to simulate the conditions of typical high-rise buildings with a slenderness ratio of 5.0. The proposed model is trained and verified by using the wind data such as wind speeds and directions and the corresponding strains measured with fiber optic grating sensors. In addition to estimation of the maximum and minimum values of strains in vertical members in a high-rise building, it is found that the proposed model can build a relationship between the wind data and strain of vertical members.     

一种多尺度卷积神经网络的人脸检测模型

卷积神经网络在检测不同尺度的人脸时所需要的计算量很大,检测过程由多个分离的步骤组成,过于复杂。针对这两方面的不足,提出一种多尺度卷积神经网络模型。根据卷积神经网络各个层具有大小不同的感受野,从不同层提取多个尺度的特征向量分别进行人脸分类与回归,并将网络的全连接层改成卷积层,以适应不同大小的图片输入。该方法将人脸检测的多个步骤集成到一个卷积神经网络中,降低了模型复杂度。实验结果表明,相同测试条件下,所提方法相比其他人脸检测模型在准确率和检测速度上均有显著提升。     

&thetas;-adaptive neural networks: a new approach to parameterestimation

A novel use of neural networks for parameter estimation in nonlinear systems is proposed. The approximating ability of the neural network is used to identify the relation between system variables and parameters of a dynamic system. Two different algorithms, a block estimation method and a recursive estimation method, are proposed. The block estimation method consists of the training of a neural network to approximate the mapping between the system response and the system parameters which in turn is used to identify the parameters of the nonlinear system. In the second method, the neural network is used to determine a recursive algorithm to update the parameter estimate. Both methods are useful for parameter estimation in systems where either the structure of the nonlinearities present are unknown or when the parameters occur nonlinearly. Analytical conditions under which successful estimation can be carried but and several illustrative examples verifying the behavior of the algorithms through simulations are presented.     

BP网络模型对人脸朝向的识别

研究BP神经网络模型,通过计算机模拟人脑建立神经元网络,使用一部分人脸朝向信息作为训练的实例集,训练稳健后,可以推广应用判断其他人脸的朝向信息,以实现计算机自动识别人脸朝向。先对图片进行归一化等预处理,再应用主成分分析提取特征信息,每幅提取出的特征信息都是8个数据的列向量,构建一个8个输入、17个隐含、3个输出的三层BP网络模型。将训练实例集的特征向量代入训练,调整参数后保证其性能和收敛速度。最后通过大量实验验证,计算机识别的误判率仅为6.7%,模型可靠。     

An automotive generator fault diagnosis system using discrete wavelet transform and artificial neural network

This paper describes a fault diagnosis system for automotive generators using discrete wavelet transform (DWT) and an artificial neural network. Conventional fault indications of automotive generators generally use an indicator to inform the driver when the charging system is malfunction. But this charge indicator tells only if the generator is normal or in a fault condition. In the present study, an automotive generator fault diagnosis system is developed and proposed for fault classification of different fault conditions. The proposed system consists of feature extraction using discrete wavelet analysis to reduce complexity of the feature vectors together with classification using the artificial neural network technique. In the output signal classification, both the back-propagation neural network (BPNN) and generalized regression neural network (GRNN) are used to classify and compare the synthetic fault types in an experimental engine platform. The experimental results indicate that the proposed fault diagnosis is effective and can be used for automotive generators of various engine operating conditions.     

A wavelet neural network for the detection of heart valve diseases

An expert system is presented for interpretation of the Doppler signals of heart valve diseases based on pattern recognition. We deal in particular with the combination of feature extraction and classification from measured Doppler signal waveforms at the heart valve using Doppler ultrasound. A wavelet neural network model developed by us is used. The model consists of two layers: a wavelet layer and a multilayer perceptron. The wavelet layer used for adaptive feature extraction in the time–frequency domain is composed of wavelet decomposition and wavelet entropy. The multilayer perceptron used for classification is a feedforward neural network. The performance of the developed system has been evaluated in 215 samples. The test results show that this system is effective to detect Doppler heart sounds. The classification rate averaged 91% correct for 123 test subjects.     

图网络层级信息挖掘分类算法综述

深度学习作为人工智能的一个研究分支发展迅速,而研究数据主要是语音、图像和视频等,这些具有规则结构的数据通常在欧氏空间中表示。然而许多学习任务需要处理的数据是从非欧氏空间中生成,这些数据特征和其关系结构可以用图来定义。图卷积神经网络通过将卷积定理应用于图,完成节点之间的信息传播与聚合,成为建模图数据一种有效的方法。尽管图卷积神经网络取得了巨大成功,但针对图任务中的节点分类问题,由于深层图结构优化的特有难点——过平滑现象,现有的多数模型都只有两三层的浅层模型架构。在理论上,图卷积神经网络的深层结构可以获得更多节点表征信息,因此针对其层级信息进行研究,将层级结构算法迁移到图数据分析的核心在于图层级卷积算子构建和图层级间信息融合。本文对图网络层级信息挖掘算法进行综述,介绍图神经网络的发展背景、存在问题以及图卷积神经网络层级结构算法的发展,根据不同图卷积层级信息处理将现有算法分为正则化方法和架构调整方法。正则化方法通过重新构建图卷积算子更好地聚合邻域信息,而架构调整方法则融合层级信息丰富节点表征。图卷积神经网络层级特性实验表明,图结构中存在层级特性节点,现有图层级信息挖掘算法仍未对层级特性节点的图信息进行完全探索。最后,总结了图卷积神经网络层级信息挖掘模型的主要应用领域,并从计算效率、大规模数据、动态图和应用场景等方面提出进一步研究的方向。     

Neural network control of nonlinear dynamic systems using hybrid algorithm

In this paper, a hybrid method is proposed to control a nonlinear dynamic system using feedforward neural network. This learning procedure uses different learning algorithm separately. The weights connecting the input and hidden layers are firstly adjusted by a self organized learning procedure, whereas the weights between hidden and output layers are trained by supervised learning algorithm, such as a gradient descent method. A comparison with backpropagation (BP) shows that the new algorithm can considerably reduce network training time.     

基于单列多尺度卷积神经网络的人群计数

单张图片和监控视频中的人群计数问题在近年来受到了越来越多的关注。尺度的变化和人群遮挡等问题,导致人群计数是一项十分具有挑战性的任务,但是深度卷积神经网络被证明能有效地解决这一问题。文中提出了一种单列多尺度的卷积神经网络,该网络提供了一种数据驱动的深度学习方法,能够理解各种不同的场景,并能进行精确的计数估计。该网络模型主要由作为二维特征提取的前端与中端,和用来还原密度图的后端组成。其中,使用堆叠池代替最大池化层,在不引入额外参数的前提下增加了模型的尺度不变性。网络模型前端采用部分VGG-16结构;中端采用FME(特征聚合模块),用来打破不同列之间的独立,以更好地提取多尺度特征信息;后端采用3列5层的不同扩张率的空洞卷积,在保持分辨率不变的情况下增加感受野,生成更高质量的人群密度图,并引入一种相对人数损失,以提升稀疏密度人群情况下模型的性能。该模型在两个最具挑战性的人群计数数据集上都取得了很好的效果。实验结果表明,在公开人群计数数据集ShanghaiTech的两个子集和UCF_CC_50上,该方法的平均绝对误差(MAE)和均方误差(MSE)分别是66.2和103.0、8.7和13.4、251.0和329.5,性能比传统人群计数方法更好。与其他模型相比,该模型拥有更高的精度和更好的鲁棒性,对稀疏人数图像有着更好的计数效果。     

基于模糊神经网络的煤层冲击地压预测模型研究

煤层冲击地压是煤矿重大灾害之一。冲击地压的发生是由多方面因素造成的,具有模糊性、动态性,表现为一个复杂的非线性动力学过程,这使得冲击地压预测系统的数据处理不能按照常规的线性系统法进行处理。文章提出了多源信息融合的模糊神经元网络算法,且基于势场拓扑层次聚类融合FCM算法的聚类思想,将模糊集合理论引入神经元网络,构成基于多判据信息融合的模糊神经元网络模型,并对该网络进行了优化。通过仿真试验,验证了该模型的有效性。     

复杂噪声环境下语音识别研究

语音识别是人机交互的重要方式,针对传统语音识别系统对含噪语音识别性能较差、特征选择不恰当的问题,提出一种基于迁移学习的深度自编码器循环神经网络模型。该模型由编码器、解码器以及声学模型组成,其中,声学模型由堆栈双向循环神经网络构成,用于提升识别性能;编码器和解码器均由全连接层构成,用于特征提取。将编码器结构及参数迁移至声学模型进行联合训练,在含噪Google Commands数据集上的实验表明本文模型有效增强了含噪语音的识别性能,并且具有较好的鲁棒性和泛化性。     

基于Elman网络的时延预测及其改进

分析了网络传输时延的组成和特点,提出了利用Elman神经网络预测网络传输时延,运用Matlab软件对其预测进行仿真,结果证明Elman神经网络能很好地预测网络时延,为了进一步提高神经网络的逼近能力和动态特性,提出了一种改进的基于输入层、隐藏层、输出层神经元的动态递归神经网络。实验证明,改进的Elman神经网络比原来的网络具有更好的动态性能。     

Nonlinear dynamic system identification using pipelined functional link artificial recurrent neural network

A computationally efficient pipelined functional link artificial recurrent neural network (PFLARNN) is proposed for nonlinear dynamic system identification using a modification real-time recurrent learning (RTRL) algorithm in this paper. In contrast to a feedforward artificial neural network (such as a functional link artificial neural network (FLANN)), the proposed PFLARNN consists of a number of simple small-scale functional link artificial recurrent neural network (FLARNN) modules. Since those modules of PFLARNN can be performed simultaneously in a pipelined parallelism fashion, this would result in a significant improvement in its total computational efficiency. Moreover, nonlinearity of each module is introduced by enhancing the input pattern with nonlinear functional expansion. Therefore, the performance of the proposed filter can be further improved. Computer simulations demonstrate that with proper choice of functional expansion in the PFLARNN, this filter performs better than the FLANN and multilayer perceptron (MLP) for nonlinear dynamic system identification.     

A multilayer incremental neural network architecture for classification

A new multilayer incremental neural network (MINN) architecture and its performance in classification of biomedical images is discussed. The MINN consists of an input layer, two hidden layers and an output layer. The first stage between the input and first hidden layer consists of perceptrons. The number of perceptrons and their weights are determined by defining a fitness function which is maximized by the genetic algorithm (GA). The second stage involves feature vectors which are the codewords obtained automaticaly after learning the first stage. The last stage consists of OR gates which combine the nodes of the second hidden layer representing the same class. The comparative performance results of the MINN and the backpropagation (BP) network indicates that the MINN results in faster learning, much simpler network and equal or better classification performance.     

Identification of complex shapes using a self organizing neuralsystem

We present a multilayer hierarchical neural system for automatic classification of complex contour patterns. The system consists of a neocognitron-like network structure combined with self-organizing maps to automatically determine feature classes. We present results showing that multilayer hierarchical networks are able to tolerate pattern distortion considerably better than standard neural network implementations.     

基于自结构动态递归模糊神经网络的无人机姿态控制*

针对无人机非线性、强耦合等特点,提出了基于该自结构动态递归模糊神经网络的姿态控制系统,给出了基于Lyapunov函数的系统稳定性证明。对四层模糊神经网络进行了优化和改进,设计了自结构动态递归模糊神经网络,该网络可以根据系统状态在线更新权值、创建/删除节点、优化网络结构。仿真表明：该控制方法的突出优点是,在兼顾考虑了系统中的不确定性因素、非线性因素及外部干扰并存的情况下,保证系统的稳定性和跟踪性能;同时此网络结构比固定结构的模糊神经网络响应速度快,因此更具优越性。     

Driver identification using finger-vein patterns with Radon transform and neural network

A driver identification system using finger-vein technology and an artificial neural network is presented in this paper. The principle of the proposed system is based on the function of near infra-red finger-vein patterns for biometric authentication. Finger-vein patterns are required by transmitting near infra-red through a finger and capturing the image with an infra-red CCD camera. The algorithm of the proposed system consists of a combination of feature extraction using Radon transform and classification using the neural network technique. The Radon transform can concentrate the information of an image in a few high-valued coefficients in the transformed domain. The neural networks are used to develop the training and testing modules. The artificial neural network techniques using radial basis function network and probabilistic neural network are proposed to develop a driver identification system. The experimental results indicated the proposed system performs well for personal identification. The average identification rate of PNN network is over 99.2%. The details of the image processing technique and the characteristic of system are also described in this paper.     

基于深度卷积神经网络End-to-End模型的亲属关系认证算法

针对如何利用人脸图像进行亲属关系认证的问题,提出基于深度卷积神经网络End-to-End模型的亲属关系认证算法.首先,构建一个包含卷积层、全连接层和soft-max分类层的深度卷积神经网络模型.卷积层可以提取亲子图像的隐性特征,全连接层可以将提取的隐性特征映射为亲属关系认证的二分类问题,soft-max分类层可以直接判断该对样本是否具有亲属关系.然后,将成对的标记训练数据输入网络进行迭代,优化深度网络模型参数,直至损失曲线稳定.最后,利用训练完毕的深度网络模型对输入测试图像对进行分类判决,通过统计得到最终的准确率.在KinFaceWI和KinFaceWII数据库上的结果显示,相比以往的亲属关系认证算法,文中模型具有更好的性能.     

一个医学图像分类器的设计

提出了一个基于径向基函数网络的医学图像分类器。该系统包括图像预处理、特征提取、分类器的构造几个部分。在网络构造上,文章采用了自适应的网络结构调整技术,提高了网络的泛化能力,同时,在网络权值调节上采用了具有全局优化的模拟退火算法,避免了陷入局部极小的缺陷。实验结果表明,该模型系统达到了76.6%的准确率,辅助系统可以极大地提高医学图像分类的效率和准确性。