Generalized clustering networks and Kohonen''s self-organizingscheme

The relationship between the sequential hard c-means (SHCM) and learning vector quantization (LVQ) clustering algorithms is discussed. The impact and interaction of these two families of methods with Kohonen's self-organizing feature mapping (SOFM), which is not a clustering method but often lends ideas to clustering algorithms, are considered. A generalization of LVQ that updates all nodes for a given input vector is proposed. The network attempts to find a minimum of a well-defined objective function. The learning rules depend on the degree of distance match to the winner node; the lesser the degree of match with the winner, the greater the impact on nonwinner nodes. Numerical results indicate that the terminal prototypes generated by this modification of LVQ are generally insensitive to initialization and independent of any choice of learning coefficient. IRIS data obtained by E. Anderson's (1939) is used to illustrate the proposed method. Results are compared with the standard LVQ approach.     

Vector quantizer of medical image using wavelet transform and enhanced SOM algorithm

Vector quantizer takes care of special image features like edges, and it belongs to the class of quantizers known as the second-generation coders. This paper proposes a novel vector quantization method using the wavelet transform and the enhanced SOM algorithm for the medical image compression. We propose the enhanced self-organizing algorithm to resolve the defects of the conventional SOM algorithm. The enhanced SOM, at first, reflects the error between the winner node and the input vector to the weight adaptation by using the frequency of the selection of the winner node. Secondly, it adjusts the weight in proportion to the present weight change and the previous one as well. To reduce the blocking effect and the computation requirement, we construct training image vectors involving image features by using the wavelet transform and apply the enhanced SOM algorithm to them for generating a well-defined codebook. Our experimental results have shown that the proposed method energizes the compression ratio and the decompression quality.     

A detection method for snoRNA modification domain by fully indexable dictionary retrieving

Recently, non-coding RNA which participates with organic activities has been found in non-coding region. Until now, we do not know detailed function of non-coding RNAs very well. To make clear functions of non-coding RNAs, we need to obtain a lot of data about non-coding RNAs and their targets. However, we do not have efficient techniques to analyze relations between non-coding RNAs and their targets. In this paper, we propose a high-speed method that can detect modification domain candidates on the target RNA based on the small nucleolar RNA (snoRNA) sequence. A snoRNA modifies a target RNA by composing complementary base pairs between a part of snoRNA and a part of target RNA. Our method stores the relations between snoRNA sequences and their target in a Fully Indexable Dictionary built by Trie data structure and Level-Order Unary Degree Sequence for high-speed retrieving.     

学习矢量量化网络在太阳耀斑预报中的应用

为了提高太阳耀斑预报的预报精度,主要是提高识别性能,提出了应用学习矢量量化(Learning Vector Quantity,LVQ)网络建立耀斑预报模型.LVQ是一种基于竞争学习规则的神经网络,采用有监督学习模式.算法每次修改连接获胜输出单元的权向鼍.更新规则是获胜单元的类别与输入向量的类别相匹配,它的权向量将向模式移近,否则将距离拉开.应用LVQ的耀斑预报模型是两层神经网络结构,输入层单元为预报因子的特征值,输出单元是耀斑预报的类别.仿真预报结果证明预撤模型具有较高的预报精度,说明LVQ是有效的预报方法.     

Counter-Propagation Neural Networks for Molecular Sequence Classification: Supervised LVQ and Dynamic Node Allocation

A modified counter-propagation (CP) algorithm with supervised learning vector quantizer (LVQ) and dynamic node allocation has been developed for rapid classification of molecular sequences. The molecular sequences were encoded into neural input vectors using an n–gram hashing method for word extraction and a singular value decomposition (SVD) method for vector compression. The neural networks used were three-layered, forward-only CP networks that performed nearest neighbor classification. Several factors affecting the CP performance were evaluated, including weight initialization, Kohonen layer dimensioning, winner selection and weight update mechanisms. The performance of the modified CP network was compared with the back-propagation (BP) neural network and the k–nearest neighbor method. The major advantages of the CP network are its training and classification speed and its capability to extract statistical properties of the input data. The combined BP and CP networks can classify nucleic acid or protein sequences with a close to 100% accuracy at a rate of about one order of magnitude faster than other currently available methods.     

基于图卷积网络和自编码器的半监督网络表示学习模型

为了保留网络结构信息和节点特征信息,结合图卷积神经网络(GCN)和自编码器(AE),提出可扩展的半监督深度网络表示学习模型(Semi-GCNAE).利用GCN捕获节点的K阶邻域中所有节点的结构和特征信息,并将捕获的信息作为AE的输入.AE对GCN捕获的K阶邻域信息进行特征提取和非线性降维,并结合Laplacian特征映射保留节点的团簇结构信息.引入集成学习方法联合训练GCN和AE,使模型习得的节点低维向量表示能同时保留网络结构信息和节点特征信息.在5个真实数据集上的广泛评估表明,文中模型习得的节点低维向量表示可以有效保留网络的结构和节点特征信息,并在节点分类、可视化和网络重构任务上性能较优.     

Feature extraction of protein expression levels based on classification of functional foods with SOM

We investigated the relations between the physiological activities and protein expression levels of functional foods using a self-organizing map (SOM). The input vectors to the SOM were the protein expression levels and the physiological activity. A competitive node has two kinds of weights: one is for protein expression levels, and the other is for physiological activity. A winner node is decided by the distance between the values of protein expression levels described in the input vector and the corresponding weights only. Then all weights, including those for physiological activity in each node, are updated. Therefore each node has an artificially generated value of physiological activity. Finally, the nodes can be categorized by the abovementioned physiological activity. A well-trained SOM gives us information about the relations between physiological activities and protein expression levels. This work was presented in part at the 13th International Symposium on Artificial Life and Robotics, Oita, Japan, January 31–February 2, 2008     

Repairs to GLVQ: a new family of competitive learning schemes

First, we identify an algorithmic defect of the generalized learning vector quantization (GLVQ) scheme that causes it to behave erratically for a certain scaling of the input data. We show that GLVQ can behave incorrectly because its learning rates are reciprocally dependent on the sum of squares of distances from an input vector to the node weight vectors. Finally, we propose a new family of models-the GLVQ-F family-that remedies the problem. We derive competitive learning algorithms for each member of the GLVQ-F model and prove that they are invariant to all scalings of the data. We show that GLVQ-F offers a wide range of learning models since it reduces to LVQ as its weighting exponent (a parameter of the algorithm) approaches one from above. As this parameter increases, GLVQ-F then transitions to a model in which either all nodes may be excited according to their (inverse) distances from an input or in which the winner is excited while losers are penalized. And as this parameter increases without limit, GLVQ-F updates all nodes equally. We illustrate the failure of GLVQ and success of GLVQ-F with the IRIS data.     

A novel feature selection approach: Combining feature wrappers and filters

Feature selection is one of the most important issues in the research fields such as system modelling, data mining and pattern recognition. In this study, a new feature selection algorithm that combines feature wrapper and feature filter approaches is proposed in order to identify the significant input variables in systems with continuous domains. The proposed method utilizes functional dependency concept, correlation coefficients and K-nearest neighbourhood (KNN) method to implement the feature filter and feature wrappers. Four feature selection methods independently select the significant input variables and the input variable combination, which yields best result with respect to their corresponding evaluation function, is selected as the winner. This is similar to the basic information fusion notion of integrating the information collected from different sources. All of the four feature selection methods are performed in two stages: (i) pre-selection, (ii) selection. Two of the four feature selection methods utilize KNN method for evaluating the candidates. These two methods use sequential forward and sequential backward search mechanism, respectively, in pre-selection stage. Whereas, the third feature selection method uses correlation coefficients in the pre-selection stage. It is common to have outliers and noise in real-life data. In order to make the proposed feature selection algorithm noise and outlier resistant, approximate functional dependencies are used by utilizing membership values that inherently cope with uncertainty in the data. Thus, the fourth feature selection method makes use of approximate functional dependencies to evaluate candidates in pre-selection stage. All of these four methods apply KNN method with exhaustive search strategy in order to find the most suitable input variable combination with respect to a performance measure.     

基于MCKD-EEMD近似熵和TWSVM的齿轮箱故障诊断

在复杂环境下齿轮箱信号往往会淹没在噪声信号中,特征向量难以提取;为了有效地进行故障诊断,提出了基于最大相关反褶积(MCKD)总体平均经验模态分解(EEMD)近似熵和双子支持向量机(TWSVM)的齿轮箱故障诊断方法;首先采用MCKD方法对强噪声信号进行滤波处理,在采用EEMD方法对齿轮箱信号进行分解,分解后得到本征模函数(IMF)分量进行近似熵求解,得到齿轮特征向量,最后将其输入到TWSVM分类器中进行故障识别;仿真实验表明,采用MCKD-EEMD方法能够有效地提取原始信号,与其他分类器相比,TWSVM的计算时间短,分类效果好等优点。     

基于DPCNN模型与语句特征融合的汉语因果类复句关系自动识别

汉语复句关系识别是对复句语义关系的识别,复句关系类别的自动识别对促进语言学和中文信息处理的研究有重要的价值。因果类复句是使用频率最高的复句,文中以二句式有标广义因果复句为研究对象, 使用语言技术平台LTP 进行依存句法分析, 获得词性、依存父节点的词序、与父节点的依存关系等特征,将特征的不同组合与预训练的词向量拼接,得到新的向量,将新的向量输入到 DPCNN 模型中来进行关系类别识别。通过实验对提出的方法进行检验,实验结果显示: 与未融合任何特征相比,DPCNN模型中融合语句特征使实验结果的指标均有提升,表明融合语句特征能取得更好的识别效果。在各种特征组合中,融合POS特征组合得到的准确度和F₁值最高, 分别为98.41%, 98.28%。     

非线性模拟电路故障诊断的小波领袖多重分形分析方法

针对非线性模拟电路故障的复杂性和非线性,提出一种基于小波领袖多重分形分析和支持向量机的故障诊断方法.首先,采用小波领袖方法对从测试节点采集的信号进行多重分形分析,并将提取的多重分形谱特征构成特征集;然后将特征集输入支持向量机,利用支持向量机的分类功能对电路的模式状态作出决策;最后,视频放大器电路故障诊断实验验证了该方法的有效性.     

基于小波变换和支持向量机的人脸检测

提出一种新的人脸检测方法——基于小波变换和支持向量机的方法。其方法的新颖之处体现在：通过综合原输人图像的小波变换值、灰度值的投影来进行特征分析;运用统计模型来估计类条件概率密度函数;运用最优的分类方法——贝叶斯分类器进行判决分类。人脸类采用正态分布建模,而非人脸类(包括除人脸类之外的一切事物)仍用正态分布来建模是不合理的。但可以用支持向量机方法从非人脸类中抽出一些跟人脸类很接近的非人脸类的特殊子集,然后对这特殊子集用正态分布进行建模。     

Multiple descent cost competition: restorable self-organization andmultimedia information processing

Multiple descent cost competition is a composition of learning phases for minimizing a given measure of total performance, i.e., cost. In the first phase of descent cost learning, elements of source data are grouped. Simultaneously, a weight vector for minimal learning, (a winner), is found. Then, the winner and its partners are updated for further cost reduction. Therefore, two classes of self-organizing feature maps are generated: a grouping feature map, and an ordinary weight vector feature map. The grouping feature map, together with the winners, retains most of the source data information. This feature map is able to assist in a high quality approximation of the original data. In the paper, the total algorithm of the multiple descent cost competition is explained and image processing concepts are introduced. A still image is first data-compressed, then a restored image is morphed using the grouping feature map by receiving directions given by an external intelligence. Next, an interpolation of frames is applied in order to complete animation coding. Examples of multimedia processing on virtual digital movies are given.     

Feature extraction and dimensionality reduction algorithms and their applications in vowel recognition

Feature extraction is an important component of a pattern recognition system. It performs two tasks: transforming input parameter vector into a feature vector and/or reducing its dimensionality. A well-defined feature extraction algorithm makes the classification process more effective and efficient. Two popular methods for feature extraction are linear discriminant analysis (LDA) and principal component analysis (PCA). In this paper, the minimum classification error (MCE) training algorithm (which was originally proposed for optimizing classifiers) is investigated for feature extraction. A generalized MCE (GMCE) training algorithm is proposed to mend the shortcomings of the MCE training algorithm. LDA, PCA, and MCE and GMCE algorithms extract features through linear transformation. Support vector machine (SVM) is a recently developed pattern classification algorithm, which uses non-linear kernel functions to achieve non-linear decision boundaries in the parametric space. In this paper, SVM is also investigated and compared to linear feature extraction algorithms.     

基于SSD的轻量级车辆检测网络改进

针对嵌入式摄像设备在执行目标检测任务过程中,对于移动中车辆的检测耗时较长无法及时反馈检测结果的问题,提出了一种基于残差连接和注意力机制的轻量级卷积网络来对SSD（single shot multibox detector）目标检测模型进行改进。采用h-swish和h-sigmoid激活函数分别替换残差块中的ReLU激活函数和通道注意力模块中的sigmoid激活函数,降低训练和推理所需计算量。根据实际应用中特定角度下车辆外形的特征为依据,重新设计SSD目标检测方法的默认框生成比例,并结合输入图像大小及特征图感受野来减少特征融合层及默认框匹配运算量。实验表明改进后的SSD检测模型在BIT-Vehicle Dataset上的平均精度均值（mean average precision,mAP）达到了94.87%,相较于经典SSD目标检测模型的mAP提升了0.83个百分点,在搭载了Intel NCS2的Raspbery PI 3+上平均处理速度达到了16?frame/s。     

Automatic RNA virus classification using the Entropy-ANFIS method

Innovations in the fields of medicine and medical image processing are becoming increasingly important. Historically, RNA viruses produced in cell cultures have been identified using electron microscopy, in which virus identification is performed by eye. Such an approach is time consuming and depends on manual controls. Moreover, detailed knowledge about RNA viruses is required. This study introduces the Entropy-Adaptive Network Based Fuzzy Inference System (Entropy-ANFIS method), which can be used to automatically detect RNA virus images. This system consists of four stages: pre-processing, feature extraction, classification and testing the Entropy-ANFIS method with respect to the correct classification ratio. In the pre-processing stage, a center-edge changing method is used, in which the Euclidian distances are calculated from the center pixels to the edges of the imaged object. In this way, the distance vector is obtained. This calculation is repeated for each RNA virus image. In feature extraction, stage norm entropy, logarithmic energy and threshold entropy values are calculated to form the feature vector. The obtained feature vector is independent of the rotation and scale of the RNA virus image. In the classification stage, the feature vector is given as input to the ANFIS classifier, ANN classifier and FCM cluster. Finally, the test stage is performed to evaluate the correct classification ratio of the Entropy-ANFIS algorithm for the RNA virus images. The correct classification ratio has been determined as 95.12% using the proposed Entropy-ANFIS method.     

Visualization of randomly ordered numeric data sets using spherical self-organizing feature maps

Scientific data visualization requires a variety of mathematical techniques to transform multivariate data sets into simple graphical objects, or glyphs, that provide scientists and engineers with a clearer understanding of the underlying system behaviour. The spherical self-organizing feature map (SOFM) described in this paper exploits an unsupervised clustering algorithm to map randomly organized N-dimensional data into a lower three-dimensional (3D) space for visual pattern analysis. Each node on the spherical lattice corresponds to a cluster of input vectors that lie in close spatial proximity within the original feature space, and neighbouring nodes on the lattice represent cluster centres with a high degree of vector similarity. Simple metrics are used to extract associations between the cluster units and the input vectors assigned to them. These are then graphically displayed on the spherical SOFM as either surface elevations or colourized facets. The resulting colourized graphical objects are displayed and manipulated within 3D immersive virtual reality (IVR) environments for interactive data analysis. The ability of the proposed algorithm to transform arbitrarily arranged numeric strings into unique, reproducible shapes is illustrated using chaotic data generated by the Lozi, Hénon, Rössler, and Lorenz attractor functions under varying initial conditions. Implementation of the basic data visualization technique is further demonstrated using the more common Wisconsin breast cancer data and multi-spectral satellite data.     

Neural-Fuzzy Models for Multispectral Image Analysis

In this paper, we consider neural-fuzzy models for multispectral image analysis. We consider both supervised and unsupervised classification. The model for supervised classification consists of six layers. The first three layers map the input variables to fuzzy set membership functions. The last three layers implement the decision rules. The model learns decision rules using a supervised gradient descent procedure. The model for unsupervised classification consists of two layers. The algorithm is similar to competitive learning. However, here, for each input sample, membership functions of output categories are used to update weights. Input vectors are normalized, and Euclidean distance is used as the similarity measure. In this model if the input vector does not satisfy the similarity criterion, a new cluster is created; otherwise, the weights corresponding to the winner unit are updated using the fuzzy membership values of the output categories. We have developed software for these models. As an illustration, the models are used to analyze multispectral images.     

Fuzzy algorithms for learning vector quantization

This paper presents the development of fuzzy algorithms for learning vector quantization (FALVQ). These algorithms are derived by minimizing the weighted sum of the squared Euclidean distances between an input vector, which represents a feature vector, and the weight vectors of a competitive learning vector quantization (LVQ) network, which represent the prototypes. This formulation leads to competitive algorithms, which allow each input vector to attract all prototypes. The strength of attraction between each input and the prototypes is determined by a set of membership functions, which can be selected on the basis of specific criteria. A gradient-descent-based learning rule is derived for a general class of admissible membership functions which satisfy certain properties. The FALVQ 1, FALVQ 2, and FALVQ 3 families of algorithms are developed by selecting admissible membership functions with different properties. The proposed algorithms are tested and evaluated using the IRIS data set. The efficiency of the proposed algorithms is also illustrated by their use in codebook design required for image compression based on vector quantization.