Community detection in complex networks by density-based clustering

We proposed a method to find the community structure in a complex network by density-based clustering. Physical topological distance is introduced in density-based clustering for determining a distance function of specific influence functions. According to the distribution of the data, the community structures are uncovered. The method keeps a better connection mode of the community structure than the existing algorithms in terms of modularity, which can be viewed as a basic characteristic of community detection in the future. Moreover, experimental results indicate that the proposed method is efficient and effective to be used for community detection of medium and large networks.     

多指标面板数据融合聚类分析

针对多指标面板数据的样品分类和历史时期划分问题,从多元统计分析理论角度提出一个多指标面板数据的融合聚类分析方法。该方法改进了多指标面板数据的因子分析和系统聚类方法,依据Fisher有序聚类理论,构造了Frobenius范数形式的离差平方和函数,提出了多指标面板数据的有序聚类方法。实证结果表明,该方法能够满足系统分析的统一性要求,保证指标之间的不相关;能够克服时间维度上均值处理造成的偏误,信息损失较少;能够解决面板数据有序聚类的问题;弥补了单一分析的片面性和局限性。     

A Clustering Approach for the Separation of Touching Edges in Particle Images

The occurrence of touching objects in images of particulate systems is very common especially in the absence of dispersion methods during image acquisition. The separation of these touching particles is essential before accurate estimation of particle size and shape can be achieved from these images. In the current work, clustering approaches based on the fuzzy C‐means algorithm are employed to identify the touching particle regions. Firstly, clustering in the multidimensional space of image features, e.g., standard deviation, gradient and range calculated in a certain neighborhood of each pixel, is performed to trap the touching regions. Then, in a novel proposed method, the clustering of pixel intensity itself into two fuzzy clusters is performed and a feature, referred to as the ‘Fuzzy Range', is calculated for each pixel from its membership values in both clusters and is presented as a distinguishing feature of the touching regions. Both approaches are compared and the superiority of the latter method in terms of the non‐necessity of neighborhood based calculations and minimum disfiguration is elucidated. The separation methods presented herein do not make any assumption about the shape of the particle as is undertaken in many methods reported elsewhere. The technique is proven to minimize greatly the deleterious effects of over‐segmentation, as is the case with traditional watershed segmentation techniques, and consequently, it results in a superior performance.     

XPS study of the SnO2 films modified with Rh

In this paper, we have analyzed the effect of the rhodium surface modification on the surface state of SnO₂ films. SnO₂ films, subjected for the surface modification, were deposited by spray pyrolysis, while Rh was deposited by using a microelectron beam evaporation. The thickness of the Rh coating varied in the range 0 to 0.1 monolayer. An explanation of the observed effects was proposed. Basing on the results of X‐ray photoelectron spectroscopy, it was assumed that at a small thickness of the rhodium covering, Rh was in a the well‐dispersed state, close to atomically dispersed state. The growth in the size of the nanoparticles began mainly when the thickness of the Rh covering exceeeded 0.01 monolayer. The size of clusters did not exceed 0.5 to 1.0 nm.     

Clustering Multivariate Longitudinal Observations: The Contaminated Gaussian Hidden Markov Model

The Gaussian hidden Markov model (HMM) is widely considered for the analysis of heterogenous continuous multivariate longitudinal data. To robustify this approach with respect to possible elliptical heavy-tailed departures from normality, due to the presence of outliers, spurious points, or noise (collectively referred to as bad points herein), the contaminated Gaussian HMM is here introduced. The contaminated Gaussian distribution represents an elliptical generalization of the Gaussian distribution and allows for automatic detection of bad points in the same natural way as observations are typically assigned to the latent states in the HMM context. Once the model is fitted, each observation has a posterior probability of belonging to a particular state and, inside each state, of being a bad point or not. In addition to the parameters of the classical Gaussian HMM, for each state we have two more parameters, both with a specific and useful interpretation: one controls the proportion of bad points and one specifies their degree of atypicality. A sufficient condition for the identifiability of the model is given, an expectation-conditional maximization algorithm is outlined for parameter estimation and various operational issues are discussed. Using a large-scale simulation study, but also an illustrative artificial dataset, we demonstrate the effectiveness of the proposed model in comparison with HMMs of different elliptical distributions, and we also evaluate the performance of some well-known information criteria in selecting the true number of latent states. The model is finally used to fit data on criminal activities in Italian provinces. Supplementary materials for this article are available online     

基于Tucker3分解的三路数据聚类方法

从两路数据聚类分析到三路数据聚类分析实质上是由平面分析到立体分析的过程。三路数据聚类方法研究的核心之一是如何把传统的两路截面数据聚类技术向三路数据聚类扩展的问题。本文基于Tucker模型的思路,提出一种先对三路数据执行矩阵分解,而后进行聚类分析的三路数据聚类方法。这种方法不但能够通过核心矩阵反映三路数据三个模式信息联系的强度大小,而且还可以在一个分解框架下对三路数据的三个模式同时进行聚类分析。实证分析结果表明,本文提出的聚类方法不但灵活、易于理解,同时也有着良好的判别性和实用性。     

灰色聚类白化函数探析

为进一步探析各类白化函数对综合评价结果的影响,先构建经典三角、修正三角1、修正三角2、经典聚类和指数型这五种不同类型的白化函数,再分别应用单一型数据、混合型数据、单指标单调型数据和跳跃型数据对这五种白化函数进行比较验证.结果显示两端等级的白化函数、零权重问题和等级区间宽度对综合评价结果影响较大.     

Dynamical characteristics of bacteria clustering by self-generated attractants

Motivated by the recent work on Escherichia coli bacteria clustering [Park, S., Wolanin, P.M., Yuzbashyan, E.A., Lin, H., Darnton, N.C., Stock, J.B., Silberzan, P., Austin, R., 2003. Proc. Natl. Acad. Sci. U.S.A. 100 (24), 13910], we have conducted a computer simulation of E. coli chemotaxis induced by a self-excreted attractant and investigated how bacteria clusters interact through a self-excreted attractant. By modeling the variation of tumbling frequency in the context of phosphorylation rate change, we have investigated the dependency of clustering behavior on the sensitivity of cells to the attractant. We have found that there exists an optimal sensitivity leading to bigger clusters and that the geometry surrounding the cells also plays an important role in localizing the cluster formation. This result suggests that bacterial cluster formation can be reduced by making bacteria more sensitive to attractants, which is opposite to an instinctive way (making them retarded to attractants). In addition, we have studied the effect of an initial cell distribution on clustering.     

Anomaly detection in a mobile communication network

Mobile communication networks produce massive amounts of data which may be useful in identifying the location of an emergency situation and the area it affects. We propose a one pass clustering algorithm for quickly identifying anomalous data points. We evaluate this algorithm’s ability to detect outliers in a data set and describe how such an algorithm may be used as a component of an emergency response management system.

Variable neighbourhood search for colour image quantization

^** Email: nenad.mladenovic{at}brunel.ac.uk Colour image quantization is a data compression technique thatreduces the total set of colours in a digital image to a representativesubset. This problem is first expressed as a large M-medianone. The advantages of this model over the usual minimum sum-of-squaresmodel are discussed first and then, the heuristic based on variableneighbourhood search metaheuristic is applied to solve it. Computationalexperience proves that this approach compares favourably withtwo other recent state-of-the-art heuristics, based on geneticand particle swarm searches.