保障用户认证安全的判别最大熵语音识别机制

为了提高电子商务中用户认证的安全性,提出了一种高识别率的判别最大熵语音识别机制DME.该语音识别方法同时考虑语音与语言两方面的因素,并将语音和语言特征进行有效地结合,在统一的最大熵模型下,实现判别训练,确保观察样本能正确地分配到其对应状态,以提高所训练出的语音模型的正确识别率.详细的实验及与现有方法的比较结果表明,对不同环境下的语音数据,提出的语音识别方法具有更好的识别性能,对提高电子商务中用户认证系统的安全性能具有理论与实际意义.     

基于深度学习的人体行为识别算法

目前,针对深度学习的人体行为识别研究,往往采用视频中的全局信息对人体行为进行分析.然而,局部信息缺失造成的特征提取不完备,同样会导致识别精度急剧下降.由此,提出了基于多流深度学习的人体行为识别方法,将人体局部信息与全局信息相结合,通过局部不同特征的精确识别,使人体行为识别更加准确.实验表明,与现有深度学习方法相比,提出的方法在数据集UCF101和HMDB51上识别精度分别平均提高了约4.0%和6.2%.     

基于Vague集模糊推理的目标识别方法及其三Ⅰ解

为从Vague集多准则模糊决策、目标识别和模糊推理三者关系中探求目标识别构建方法,利用特征矩阵、权重、评价函数等分别构建了基本决策过程和熵权-加权算子决策过程.通过R_(0v)。型模糊取式三Ⅰ算法,揭示了两种决策过程的评价函数值分别是模糊推理的模糊逻辑三Ⅰ解和加权三Ⅰ解.结合三Ⅰ算法的还原性和模糊推理过程,构建了基于三Ⅰ算法的模糊推理目标识别方法.利用一个工件识别实例说明了提出的目标识别方法的正确性和有效性.     

基于sketch数据结构与正则性分布的骨干网流量异常分析与识别

随着互联网爆炸性的增长,网络安全问题如恶意攻击,蠕虫病毒,DDoS攻击等事件的数量和影响也一直在不断增加.如何能够及时准确地检测到网络流量异常是我们面临的一个重要问题.作者根据网络流量的信号特性,结合sketch数据结构和网络流量Lipschitz正则性分布,提出了一种新的异常检测技术.该方法不仅能有效地定位骨干网流量异常发生的源IP地址和发生时刻,还能根据源IP地址熵值的分析,对异常进行识别。实验分析的结果验证了该方法在检测和可溯源方面的有效性.     

基于BP神经网络的手写体数字的识别方法研究

对手写体数字的识别问题进行了讨论,提出一种基于BP神经网络的识别方法.从而提高了识别效率.主要就在识别时,数字在图片上的位置和数字本身大小方面做了改进,发现数字在图片上的大小和其在图片上的位置直接影响识别效果.具体做的是,首先提取了图片的轮廓,然后归一化成28×28的图像.这样做,不仅使得图像数字区域大小相同,而且都在图像中心上,使得识别结果变的更加理想化,达到了高识别的目的.另外,选择了容错性较好的BP网络,以200组手写体数字图像作为输入向量,以其他的110组进行识别,效率达到了90%.     

空间一维结构分布动载荷时域识别方法

提出一种适用于空间一维结构分布动载荷的时域识别方法.基于空间分段和时间分段的思想, 推导了载荷时程识别的公式和过程.用MATLAB编写了载荷识别程序,以受分布动载荷简支梁和受随机风载荷输电导线的载荷识别对方法和程序进行验证, 并通过仿真试验研究了噪声对载荷识别的影响.结果表明,该方法对于线性问题有很高的识别精度,对于弱非线性问题能够满足工程应用要求,为分布动载荷时域识别提供了有效的途径.     

复杂背景下运动车辆车标定位与识别方法

目前关于运动车辆识别技术主要为基于图像处理的车牌自动识别技术,该技术主要适用于在简单背景下运行的单一车辆的牌照识别,对于在复杂背景下运行以及车牌被更换的运动车辆的车牌识别误差较大.为了弥补车牌识别技术在进行车辆识别时的不足,提出了一种复杂背景下运动车辆车标定位与识别方法.     

基于IFOA-ELM的车载手势识别方法研究

车载手势识别作为一种人机交互方式是提高道路行驶安全性的有效途径.针对传统车载手势识别方法研究中识别准确率和效率较低以及性能不稳定的问题,提出一种改进的果蝇优化算法(IFOA)优化极限学习机(ELM)参数的车载手势识别新方法.首先利用IFOA优化ELM的初始权重w和偏置b;接着采用最佳初始权重和偏置来训练ELM;最后利用IFOA-ELM对提取的车载手势特征向量进行手势类型识别;实验结果表明,与SVM、动态贝叶斯网络、传统ELM、FOA-ELM等分类学习算法相比,方法在高效稳定的前提下取得更高的识别准确率,满足对准确性和实时性要求较高的车载环境中的手势识别.     

应用RJMCMC方法识别线性模型异常点

传统线性模型异常点识别方法容易发生误判:正常点被归为异常点或者异常点被归为正常点.为解决此类问题,提出了应用逆跳马尔科夫蒙特卡洛方法识别异常点的思想,同时将其应用于实际数据加以检验,识别效果明显好于传统方法.     

基于DEA的竞争战略识别

针对有效决策单元评价和区分的问题,在充分提取决策单元之间相似性和相异性信息基础上,定义了多指标区间交叉效率,进而提出了一种基于投入、产出权重的聚类分析方法,并将其应用于竞争战略识别.实证结果表明,该方法能够区分有效决策单元,综合评价具有统一性和合理性;与同类战略识别方法相比,更具客观性和解释能力,分类效果更好.该方法提供了一种客观的新的竞争战略识别方法,有助于战略有效性的研究.     

Reporting a network’s most-central actor with a confidence level

This article introduces a confidence level (CL) statistic to accompany the identification of the most central actor in relational, social network data. CL is the likelihood that the most-central actor assertion is correct in light of imperfect network data. The CL value is derived from a frequency-based probability according to perturbed samples of feature-equivalent network data. Analysts often focus attention towards the most central, highest valued, top actor [or node] according to one of four traditional measures: degree, betweenness, closeness or eigenvector centrality. However, given that collected social network data often has missing relational links, the correctness of the top-actor claim becomes uncertain. This paper describes and illustrates a practical approach for estimating and applying a CL to the top-actor identification task. We provide a simple example of the technique used to derive a posterior probability, then apply the same approach to larger, more pragmatic random network by using the results of an extensive virtual experiment involving uniform random and scale-free topologies. This article has implications in organizational practice and theory; it is simple and lays groundwork for developing more intricate estimates of reliability for other network measures.     

Object oriented modeling of social networks

The aim of this paper is to explain principles of object oriented modeling in the scope of modeling dynamic social networks. As such, the approach of object oriented modeling is advocated within the field of organizational research that focuses on networks.We provide a brief introduction into the field of social networks and present an overview of existing network models and methods. Subsequently we introduce an elementary problem field in the social sciences in general, and in studies of organizational change and design in particular: the micro-macro link. We argue that the most appropriate way to hadle this problem is the principle of methodological individualism. For social network analysis, to contribute to this theoretical perspective, it should include an individual choice mechanism and become more dynamically oriented. Subsequently, object oriented modeling is advocated as a tool to meet these requirements for social network analysis. We show that characteristics of social systems that are emphasized in the methodological individualistic approach have their direct equivalences in object oriented models. The link between the micro level where actors act, and the macro level where phenomena occur as a consequence and cause of these actions, can be modelled in a straightforward way.     

A novel method for the network reliability in terms of capacitated-minimum-paths without knowing minimum-paths in advance

Many real-world systems (such as electric power, transportation, etc) may be regarded as flow networks whose arcs have independent, discrete, limited and multivalued random capacities. In this study, a novel method for the network reliability is present. Analysis of the proposed algorithm and comparison to existing best-known algorithms shows that the proposed method has the following advantages: (1) it is just based on the special property of capacitated-minimum-paths (CMPs) of which the max-flow in these paths are all equal to a given capacity (say d) and can be used to search for all capacitated minimum-paths without knowing all minimum-paths in advance; (2) it is simple and more effective in finding CMP candidates than the existing methods and (3) the proposed method is easier to understand and implement.     

A mapping strategy for the identification of structural systems

In this paper an alternative approach for identification problems is discussed. Unlike existing methods, this new approach combines in a general way finite differences and function approximation and is herein used for the identification of a particular system in structural dynamics, that is the damped Duffing oscillator subject to a swept-sine excitation. The solution obtained by means of the proposed method has been compared with the one obtained by a neural network. The present method gives better results at a low computational cost, with the advantage of solutions in explicit form. Besides, it is possible to prove that the solutions are stable and that from this new approach one can deduce, as a particular case, the approximation previously proposed by other authors.     

Automatic structure and parameter training methods for modeling of mechanical systems by recurrent neural networks

Automatic nonlinear-system identification is very useful for various disciplines including, e.g., automatic control, mechanical diagnostics and financial market prediction. This paper describes a fully automatic structural and weight learning method for recurrent neural networks (RNN). The basic idea is training with residuals, i.e., a single hidden neuron RNN is trained to track the residuals of an existing network before it is augmented to the existing network to form a larger and, hopefully, better network. The network continues to grow until either a desired level of accuracy or a preset maximal number of neurons is reached. The method requires no guessing of initial weight values or the number of neurons in the hidden layer from users. This new structural and weight learning algorithm is used to find RNN models for a two-degree-of-freedom planar robot, a Van der Pol oscillator and a Mackey–Glass equation using their simulated responses to excitations. The algorithm is able to find good RNN models in all three cases.     

A computational comparison of the dinic and network simplex methods for maximum flow

We study the implementation of two fundamentally different algorithms for solving the maximum flow problem: Dinic's method and the network simplex method. For the former, we present the design of a storage-efficient implementation. For the latter, we develop a "steepest-edge" pivot selection criterion that is easy to include in an existing network simplex implementation. We compare the computational efficiency of these two methods on a personal computer with a set of generated problems of up to 4 600 nodes and 27 000 arcs.This research was supported in part by the National Science Foundation under Grant Nos. MCS-8113503 and DMS-8512277.     

Network-based naive Bayes model for social network

Naive Bayes (NB) is one of the most popular classification methods. It is particularly useful when the dimension of the predictor is high and data are generated independently. In the meanwhile, social network data are becoming increasingly accessible, due to the fast development of various social network services and websites. By contrast, data generated by a social network are most likely to be dependent. The dependency is mainly determined by their social network relationships. Then, how to extend the classical NB method to social network data becomes a problem of great interest. To this end, we propose here a network-based naive Bayes (NNB) method, which generalizes the classical NB model to social network data. The key advantage of the NNB method is that it takes the network relationships into consideration. The computational effciency makes the NNB method even feasible in large scale social networks. The statistical properties of the NNB model are theoretically investigated. Simulation studies have been conducted to demonstrate its finite sample performance. A real data example is also analyzed for illustration purpose.     

Parameter estimation and topology identification of uncertain fractional order complex networks

This paper focuses on a significant issue in the research of fractional order complex network, i.e., the identification problem of unknown system parameters and network topologies in uncertain complex networks with fractional-order node dynamics. Based on the stability analysis of fractional order systems and the adaptive control method, we propose a novel and general approach to address this challenge. The theoretical results in this paper have generalized the synchronization-based identification method that has been reported in several literatures on identifying integer order complex networks. We further derive the sufficient condition that ensures successful network identification. An uncertain complex network with four fractional-order Lorenz systems is employed to verify the effectiveness of the proposed approach. The numerical results show that this approach is applicable for online monitoring of the static or changing network topology. In addition, we present a discussion to explore which factor would influence the identification process. Certain interesting conclusions from the discussion are obtained, which reveal that large coupling strengths and small fractional orders are both harmful for a successful identification.     

Information dynamics algorithm for detecting communities in networks

The problem of community detection is relevant in many scientific disciplines, from social science to statistical physics. Given the impact of community detection in many areas, such as psychology and social sciences, we have addressed the issue of modifying existing well performing algorithms by incorporating elements of the domain application fields, i.e. domain-inspired. We have focused on a psychology and social network-inspired approach which may be useful for further strengthening the link between social network studies and mathematics of community detection. Here we introduce a community-detection algorithm derived from the van Dongen’s Markov Cluster algorithm (MCL) method [4] by considering networks’ nodes as agents capable to take decisions. In this framework we have introduced a memory factor to mimic a typical human behavior such as the oblivion effect. The method is based on information diffusion and it includes a non-linear processing phase. We test our method on two classical community benchmark and on computer generated networks with known community structure. Our approach has three important features: the capacity of detecting overlapping communities, the capability of identifying communities from an individual point of view and the fine tuning the community detectability with respect to prior knowledge of the data. Finally we discuss how to use a Shannon entropy measure for parameter estimation in complex networks.     

Solving bilinear tensor least squares problems and application to Hammerstein identification

Bilinear tensor least squares problems occur in applications such as Hammerstein system identification and social network analysis. A linearly constrained problem of medium size is considered, and nonlinear least squares solvers of Gauss–Newton‐type are applied to numerically solve it. The problem is separable, and the variable projection method can be used. Perturbation theory is presented and used to motivate the choice of constraint. Numerical experiments with Hammerstein models and random tensors are performed, comparing the different methods and showing that a variable projection method performs best.