水平吸收投影重构离散集合的快速算法

通过分析吸收系数为黄金比例的EDT物理模型,沿着左右水平吸收投影,考虑二元矩阵的重构问题及其惟一性。针对投影差值计算复杂性的不足,提出一个基于序列一致性判断条件的快速算法,将其应用于斜线吸收投影下二元矩阵的重构。与已有算法进行对比,提高了搜索解的速度。     

广义的鉴别局部中值保持投影及人脸识别

针对鉴别的局部中值保持投影(DLMPP)在小样本情况下面临的类内散布矩阵奇异的问题,提出了广义的鉴别局部中值保持投影(GDLMPP)算法。GDLMPP首先将样本等价映射到一个低维子空间,然后在此子空间求解最佳投影矩阵,从而有效解决了小样本问题,并从理论上验证了当类内散布矩阵非奇异时,GDLMPP等价于DLMPP。最后,通过在ORL及AR库上的实验验证了算法的有效性。     

对角矩阵指数优化的局部保持映射算法

局部保持映射（LPP）算法利用欧几里德距离求得权值累加得到对角矩阵,利用结果进行降维。对于这个算法是否可以进一步优化还值得进一步探讨。对该算法所依据的公式进行修改,在对角矩阵上引入指数参数,形成对角距阵指数优化的局部保持映射算法。通过实验可以证明,对角距阵指数优化的局部保持映射算法能够影响降维的结果,可以使得降维更容易得到接近本征维数的投影向量,通过实验验证降维后的识别效果和对噪声的敏感度。     

基于推广流形学习的高分辨遥感影像目标分类

针对传统的流形学习算法不能对位于黎曼流形上的协方差描述子进行有效降维这一问题，本文提出一种推广的流形学习算法，即基于Log-Euclidean黎曼核的自适应半监督正交局部保持投影（Log-Euclidean Riemannian kernel-based adaptive semi-supervised orthogonal locality preserving projection，LRK-ASOLPP），并将其成功用于高分辨率遥感影像目标分类问题.首先，提取图像每个像素点处的几何结构特征，计算图像特征的协方差描述子；其次，通过采用Log-Euclidean黎曼核将协方差描述子投影到再生核Hilbert空间；然后，基于流形学习理论，建立黎曼流形上半监督正交局部保持投影算法模型，利用交替迭代更新算法对目标函数进行优化求解，同时获得相似性权矩阵和低维投影矩阵；最后，利用求得的低维投影矩阵计算测试样本的低维投影，并用K—近邻、支持向量机（Support victor machine，SVM）等分类器对其进行分类.三个高分辨率遥感影像数据集上的实验结果说明了该算法的有效性与可行性.     

核最大方差差分嵌入在生物特征提取中的应用

最大方差差分嵌入算法（VDE）最大化全局方差和局部方差之差,该算法直接通过求解一个特征值问题而获得投影矩阵,无需矩阵求逆运算,因此VDE克服了无监督鉴别投影（UDP）算法的小样本问题,为了进一步增强VDE算法的非线性描述能力,提出了核最大方差差分嵌入算法（KVDE）,该算法首先采用核函数将样本映射到非线性高维空间,然后采用核方法得到一个低维子空间,人脸和掌纹数据库上的实验表明KVDE算法比VDE算法具有更好的性能。     

核的正交完备鉴别局部保持投影

针对完备鉴别局部保持投影算法所求得的最优判别矢量间存在信息冗余问题,提出了核的正交完备鉴别局部保持投影算法。通过将核函数技术与正交性原理融合,采用高斯核函数将原始样本映射到高维特征空间,在高维特征空间的局部总体散度矩阵中计算最优判别矢量,只需在整个范围内对值域空间进行特征值分解,去除局部零空间达到样本降维目的。该算法分别在 UMIST 人脸库和 JAFFE 人脸表情库上进行实验,实验结果表明算法的识别率高达95．59％。     

基于自适应锚定邻域回归的图像超分辨率算法

在基于字典的图像超分辨率（SR）算法中，锚定邻域回归超分辨率（ANR）算法由于其优越的重建速度和质量引起了人们的广泛关注。然而，ANR算法的锚定邻域投影并不稳定，以致于不足以涵盖各种样式的映射关系。因此提出一种基于自适应锚定邻域回归的图像SR算法，根据样本分布自适应地计算邻域中心从而以更精确的邻域来预计算投影矩阵。首先，以图像块为中心，运用K均值聚类算法将训练样本聚类成不同的簇；然后，用每个簇的聚类中心替换字典原子来计算相应的邻域；最后，运用这些邻域来预计算从低分辨率（LR）空间到高分辨率（HR）空间的映射矩阵。实验结果表明，所提算法在Set14上平均重建效果以31.56 dB的峰值信噪比（PSNR）及0.8712的结构相似性（SSIM）优于其他基于字典的先进算法，甚至胜过超分辨率卷积神经网络（SRCNN）算法。同时，在主观表现上看，所提算法恢复出了尖锐的图像边缘且产生的伪影较少。     

基于小波投影和离散哈希的图像检索

现有的哈希方法难以快速实现原始特征空间的近似映射.针对此问题,文中提出基于小波投影的哈希方法.基于Haar小波变换构造投影矩阵,使用迭代算法优化投影矩阵和离散优化二进制码,重构量化误差.利用投影矩阵将图像的原始特征向量快速投影至低维空间,并进行二进制嵌入,完成图像的哈希编码.在图像数据集上的实验表明,文中方法可有效提升编码效率.     

一种改进的ISOMAP分类算法

传统的等距特征映射算法在降维时未考虑数据的类别标签,降维后不能够产生从高维到低维的映射矩阵,且不适用于多个类簇的情况,不能直接用于分类。针对这几个问题利用近邻元分析方法取代多维尺度分析法,并且引入特征向量作为输入矩阵,提出一种以分类为目的的等距特征映射算法(NC-ISOMAP)。降维时获取理想的低维投影矩阵,使降维后类间数据更加分开,类内数据更加紧凑。实验结果表明NC-ISOMAP算法能够取得很好的降维效果和分类性能,并在不同的数据集中有着较好的鲁棒性。     

鱼眼投影在虚拟实景中的应用研究

为了生成大视野的虚拟场景扣逼真的模拟球面，本文把鱼眼投影应用到虚拟实景中，给出由经纬映射图像生成角鱼眼投影的算法，并且在实现了三维浏览扣缩放，最后通过实例证明了鱼眼投影虚拟空间的效果．     

A Dual Neural Network for Kinematic Control of Redundant Manipulators Using Input Pattern Switching

This paper presents a dual neural network for kinematic control of a seven degrees of freedom robot manipulator. The first network is a static multilayer perceptron with two hidden layers which is trained to mimic the Jacobian of a seven DOF manipulator. The second network is a recurrent neural network which is used for determining the inverse kinematics solutions of the manipulator; The redundancy is used to minimize the joint velocities in the least squares sense. Simulation results show relatively good comparison between the outputs of the actual Jacobian matrix and multilayer neural network. The first network maps motions of the seven joints of the manipulator into 42 elements of the Jacobian matrix, with surprisingly smaller computations than the actual trigonometric function evaluations. A new technique, input-pattern-switching, is presented which improves the global training of the static network. The recurrent network was designed to work with the neural network approximation of the Jacobian matrix instead of the actual Jacobian. The combination of these two networks has resulted in a time-efficient procedure for kinematic control of robot manipulators which avoids most of the complexity present in the classical-trigonometric-based methods. Also, by electronic implementation of the networks, kinematic solutions can be obtained in a very timely manner (few nanoseconds).     

Neural network-based light attenuation model for monitoring seagrass population in the Indian river lagoon

Seagrasses have been considered one of the most critical marine habitat types of coastal and estuarine ecosystems such as the Indian River Lagoon. They are an important part of biological productivity, nutrient cycling, habitat stabilization and species diversity and are the primary focus of restoration efforts in the Indian River Lagoon. The areal extent of seagrasses has declined within segments of the lagoon over the years. Light availability to seagrasses is a major criterion limiting their distribution. Decreased water clarity and resulting reduced light penetration have been cited as the major factors responsible for the decline in seagrasses in the lagoon. Hence, light is a critical factor for the survival of seagrass species. Light attenuation coefficient is an important parameter that indicates the light attenuated by the water column and can therefore be used as an indicator of seagrass vigor. A number of region-specific linear light attenuation models have been proposed in the literature. Though, in practice, linear light attenuation models have been commonly used, there is need for a flexible and robust model that incorporates the non-linearities present in coastal and estuarine environments. This paper presents a neural network based model to estimate light attenuation coefficient from water quality parameters and thereby indirectly monitor seagrass population in the Indian River Lagoon. The proposed neural network models were compared with linear regression models, step-wise linear regression models, model trees and support vector machines. The neural network models performed fairly better compared to the other models considered.     

Arbitrary FIR Filter Synthesis Using a Neural Network

We propose a neural network to synthesize an arbitrary FIR filter in a least square sense. The network can evolve to its steady-states or equilibrium points from any initial state in the magnitude of the circuit's time constant. Under the steady-state, the output of the network is just our designed FIR filter coefficient if a real, symmetric, and positive-definite matrix calculated by the design specifications is directly used as the synaptic strength matrix.     

基于径向基神经网络的电涡流传感器输出特性拟合研究

针对电涡流传感器的输出特性参数非线性较大,不能精确地反映被测物理量的问题,提出了一种采用径向基神经网络对电涡流传感器的输出特性参数进行拟合的方案。该方案采用newrb函数创建一个径向基神经网络,以被测物理量作为输入矩阵、电涡流传感器输出电压作为输出矩阵,对该径向基神经网络进行训练,从而可得到均方根误差小且光滑的电涡流传感器输出特性拟合曲线。实验结果表明,只要选择合适的创建函数和扩展系数,径向基神经网络能有效地实现电涡流传感器输出特性的拟合。     

Convolutional neural network feature maps selection based on LDA

Convolutional neural network (CNN), as widely applied to vision and speech, has developed lager and lager network size in last few years. In this paper, we propose a CNN feature maps selection method which can simplify CNN structure on the premise of stabilize the classifier performance. Our approach aims to cut the feature map number of the last subsampling layer and achieves shortest runtime on the basis of Linear Discriminant Analysis (LDA). We rebuild feature maps selection formula based on the between-class scatter matrix and within-class scatter matrix, because LDA can lead to information loss in the dimension-reduction process. Our experiments measure on two standard datasets and a dataset made by ourselves. According to the separability value of each feature map, we suggest the least number of feature maps which can keep the classifier performance. Furthermore, we prove that separability value is an effective indicator for reference to select feature maps.     

Wood defects classification using laws texture energy measures and supervised learning approach

Machine vision based inspection systems are in great focus nowadays for quality control applications. The proposed work presents a novel approach for classification of wood knot defects for an automated inspection. The proposed technique utilizes gray level co-occurrence matrix and laws texture energy measures as texture feature extractors and feed-forward back-propagation neural network as classifier. The proposed work involves the comparison of gray level co-occurrence matrix based features with laws texture energy measures based features. Firstly it takes contrast, correlation, energy and homogeneity as input parameters to a feed-forward back propagation neural network to predict wood defects and then it take energy calculated from laws texture energy measures based energy maps as input feature to a feed-forward back propagation neural network. Mean Square Error (MSE) for training data is found to be 0.0718 and 90.5% overall average classification accuracy is achieved when laws texture energy measures based features are used as input to the neural network as compared to gray level co-occurrence matrix based input features where MSE for training data is found to be 0.10728 and 84.3% overall average classification accuracy is achieved. The proposed technique shows promising results to classify wood defects using a feed forward back-propagation neural network.     

Neural network modelling on temperature coefficient of surface tension and its usage in melting point prediction of nanosized metal particles

Temperature coefficient of surface tension is a very important parameter to calculate phase diagrams of nanoparticle metal systems. In this paper, neural network calculation was for the first time used to evaluate the temperature coefficient. It shows that the constructed neural network can predict the temperature coefficient values for 37 metals, with the deviation from the averaged experimental measurements smaller than 25%. Furthermore, the neural network predictions were compared with the calculated values by using an empirical equation and it shows a better performance.     

Classification of binary document images into textual or nontextual data blocks using neural network models

This paper describes a new method for the classification of binary document images as textual or nontextual data blocks using neural network models. Binary document images are first segmented into blocks by the constrained run-length algorithm (CRLA). The component-labeling procedure is used to label the resulting blocks. The features for each block, calculated from the coordinates of its extremities, are then fed into the input layer of a neural network for classification. Four neural networks were considered, and they include back propagation (BP), radial basis functions (RBF), probabilistic neural network (PNN), and Kohonen's self-organizing feature maps (SOFMs). The performance and behavior of these neural network models are analyzed and compared in terms of training times, memory requirements, and classification accuracy. The experiments carried out on a variety of medical journals show the feasibility of using the neural network approach for textual block classification and indicate that in terms of both accuracy and training time RBF should be preferred.     

基于卷积神经网络的知识图谱补全方法研究

知识图谱是事实三元组的集合,其表示形式为(头实体,关系,尾实体)。为了补全知识图谱中缺失的实体和关系,提出一种基于卷积神经网络的知识图谱补全方法。使用传统嵌入模型训练三元组,得到实体向量和关系向量;将三元组表示成3列矩阵,作为卷积神经网络的输入,卷积后得到三元组的特征表示图;连接所有特征图和权重向量进行点乘得到每个三元组的得分,得分越低证明三元组越正确。实验采用数据集WN18RR、FB15K-237、FB15K分别进行链接预测和三元组分类实验。实验结果表明,与其他方法相比,该方法在Mean Rank和Hit@10指标上都取得了更好的实验结果,证明其可以有效提高三元组预测精度。     

A recurrent neural network for solving Sylvester equation with time-varying coefficients

Presents a recurrent neural network for solving the Sylvester equation with time-varying coefficient matrices. The recurrent neural network with implicit dynamics is deliberately developed in the way that its trajectory is guaranteed to converge exponentially to the time-varying solution of a given Sylvester equation. Theoretical results of convergence and sensitivity analysis are presented to show the desirable properties of the recurrent neural network. Simulation results of time-varying matrix inversion and online nonlinear output regulation via pole assignment for the ball and beam system and the inverted pendulum on a cart system are also included to demonstrate the effectiveness and performance of the proposed neural network.