Incorporating linguistic structure into maximum entropy language models

In statistical language models,how to integrate diverse linguistic knowledge in a general framework for long-distance dependencies is a challenging issue,In this paper,an improved language model incorporating linguistic structure into maximum entropy framework is presented.The poposed model combines trigram with the structure knowledge of base phrase in which trigram is used to capture the local relation between words.while the structure knowledge of base phrase is considered to represent the long-distance relations between syntactical structures.The knowledge of syntax,semantics and vocabulary is is integrated into the maximum entropy framework,Experimental results show that the proposed model improves by 24% for language model perplexity and increases about3% for sign language recognition rate compared with the trigram model.     

基于有理平方根B样条模型的概率密度函数形状控制

This paper presents a new method for the modeling and control of the output probability density functions (PDFs) of linear stochastic systems. At first, a new PDF approximation method, namely the rational square-root B-spline model is proposed and the innovative concept of pseudoweights is introduced. The new model is then compared with the existing B-spline models in terms of feasible domains. Next, a controller is developed to realize the output PDF tracking performance. An alternative minimal entropy control strategy is also provided for the case that no target PDF is available. Finally, illustrative examples indicate the effectiveness of the proposed algorithms.     

Shaping of Output PDF Based on the Rational Square-root B-spline Model

This paper presents a new method for the modeling and control of the output probability density functions (PDFs) of linear stochastic systems.At first,a new PDF approximation method, namely the rational square-root B-spline model is proposed and the innovative concept of pseudo- weights is introduced.The new model is then compared with the existing B-spline models in terms of feasible domains.Next,a controller is developed to realize the output PDF tracking performance. An alternative minimal entropy control strategy is also provided for the case that no target PDF is available.Finally,illustrative examples indicate the effectiveness of the proposed algorithms.     

Target models and waveform design for detection in MIMO radar

In this paper,we propose a general target model for multiple-input multiple-output(MIMO)radar.Two types of targets are presented.For each type of target,we develop a corresponding optimum waveform designing method to optimize detection performance.Based on Chernof-Stein lemma,we use relative entropy as the performance measure.We show that the optimum waveform designing method can obtain better detection performance than traditional orthogonal waveform designing method.     

Intuitionistic Logic as the Implement of Incremental Model Construction for Natural Language

The process of understanding natural language can be viewed as the process of model construction.This paper,employing Kripke frame for intuitionistic logic semantics as the implement of model construction for natural language,introduces a method of incremental model constuction.     

On detection performance of MIMO radar for Rician target

By using spatial diversity, multiple-input-multiple-output (MIMO) radar can improve detection performance for fluctuating targets. In this paper, we propose a spatial fluctuation target model for MIMO radar, where targets are classified as non-fluctuating target, Rayleigh target and Rician target. Based on Stein’s lemma, we use relative entropy to study detection performance of optimum detector for Rician target. It is found that in low signal noise ratio (SNR) region, the performance improvement of MIMO ra...     

Analysis and Applications of PCA Information Features

This paper makes a brief introduction of the principle of Principal Component Analysis （PCA）. Then according to the information entropy theory, and making full use of the inherent characteristic of eigenvalues of correlation matrix of data, the 2^nd information function, the 2nd information entropy and geometry entropy under PCA are proposed firstly, by which the information features of PCA are metricized. In addition, two new concepts of Information Rate （IR） and Accumulated Information Rate （AIR） are proposed, which are used to illustrate the degree of information feature extraction of PCA. In the end, through simulated application in practice, the results show that the method proposed in this paper is efficient and satisfactory. It provides a new research approach of information feature extraction for pattern recognition, machine learning, and data mining and so on.     

Sliding Mode Control Design via Reduced Order Model Approach

This paper presents a design of continuous-time sliding mode control for the higher order systems via reduced order model.It is shown that a continuous-time sliding mode control designed for the reduced order model gives similar performance for the higher order system.The method is illustrated by numerical examples.The paper also introduces a technique for design of a sliding surface such that the system satisfies a cost-optimality condition when on the sliding surface.     

信息论在异常检测中的应用

异常检测能有效地检测出新类型的攻击,但是目前大多数入侵检测使用的异常检测技术都缺乏理论指导,通常是经过反复的实验建立检测模型,这样建立的模型只适用于某些环境,不具有通用性。该文提出了一种可以指导异常检测的理论基础———信息论,介绍了应用在异常检测中的几种信息论的方法,包括熵,条件熵,相对(条件)熵,信息量。利用这些方法可以指导异常模型建立过程和解释模型性能。最后该文还通过举例说明了这种理论的可行性和必要性。     

Maximum Bayesian entropy method for determining ordered weighted averaging operator weights

Determination of the ordered weighted averaging (OWA) operators is an important issue in the theory of the OWA operator weights. In this paper, the main existing models for determining the OWA operator weights are outlined and the concept of the Bayesian entropy is introduced. Based upon the Bayesian entropy the maximum Bayesian entropy approach for obtaining the OWA operator weights is proposed. In this model it is assumed, according to previous experiences or from theoretical considerations that a decision maker may have reasons to consider a given prior OWA vector. Finally the new model is solved according to the prior OWA vector with specific level of orness comparing the results with other methods. The results demonstrate the efficiency of our model in generating the OWA operator weights. An applied example is also presented to illustrate the applications of the proposed model.     

信息熵在粗糙集中衍生的几个概念

随着对粗糙集理论研究的的深入,基于信息论的信息熵陆续被引入到粗糙集研究中,陆续产生了一些如条件熵、联合熵、知识熵、决策熵、知识粗糙熵、粗集粗糙熵等新的概念,尽管丰富了粗糙集理论和应用,但使用中存在语义不统一的地方,甚至缺乏必要的说明和证明。对这些有价值的新概念作了系统的、严格的、规范的定义及阐述,给出了它们的公式表示,同时,通过相关熵的运算揭示彼此间的关系,最后指出这些熵的应用范畴,以便研究人员在清楚概念的基础上作进一步研究。     

信息融合中的有关熵理论

将信息论中有关熵的概念和应用引入到目标观测识别系统中，建立了观测决策融合系统的熵模型；提出了观测决策融合系统中有关条件熵的两个重要定理，并指出它们在理论研究和实际应用中的指导意义；最后阐述了融合系统中熵的结构关系．     

新的信息熵属性约简

在粗糙集不确定性度量公式中,模糊熵和模糊度是重要的度量方式。根据粗糙集不确定性度量中模糊熵和新的模糊度公式,提出了在决策信息系统中修正条件信息熵和相对模糊熵的概念,并分别用两种方式证明了熵在属性约简过程中的单调性。然后利用向前添加属性算法进行属性约简,约简结果在RIDAS(roughset based intelligent data analysis system)平台上进行识别率测试,通过实验对比分析了两种新的信息熵与条件信息熵的约简结果,为基于信息熵的属性约简提供了参考。     

基于信息熵的风格绘画分类研究

针对艺术风格绘画分类算法中存在的精度和效率不高等问题,提出一种基于信息 熵的艺术风格绘画分类算法。首先选取西方漫画、素描、油画、水彩画,以及国内烙画、水墨 画、壁画具有代表性的 7 种艺术绘画风格作为研究对象,对图像进行去噪、归一化等预处理。 其次,提取绘画艺术作品风格特征,分别求取图像的颜色熵、分块熵、轮廓熵,并合并构成不 同绘画风格的信息熵。信息熵求取时,将色彩空间转换为 Lab 颜色空间,通过 a、b 通道颜色值 及加权函数获得图像的颜色熵;通过对艺术图像分块求取信息熵,求取分块的信息熵均值获得 分块熵;通过 Contourlet 变换,求取艺术图像的轮廓信息,获得轮廓熵。接着,合并提取的颜 色熵、分块熵、轮廓熵,利用支持向量机(SVM)对艺术风格图像学习训练,获得艺术绘画风格 的分类模型;最后,提取待识别绘画风格样本的熵特征,通过 SVM 分类识别获得最终的分类 结果。该方法具有特征维数少、运算速度快、尺度不变性等优势,实验结果表明,其能提高不 同绘画风格的分类精度和效率。     

基于联合熵的多属性匿名度量模型

提出了基于联合熵的多属性匿名度量模型，该模型基于识别性、连接性、跟踪性等匿名属性．鉴于匿名的随机性和模糊性特点，提出了基于联合熵和最小加权广义距离的模糊模式识别方法，实现了系统匿名等级隶属度向量的离散化．给出了联合熵和加权广义距离之间平衡参数的确定方法．分析表明，该模型优于现有的单属性Shannon熵模型，平衡参数是从系统的角度惟一求解．因此，联合熵可以作为匿名等级的评价指标．     

基于最大熵模型的汉语短语间停顿识别

正确标记短语间的停顿,对提高文语转换系统合成语音的自然度起着重要作用。介绍一种采用最大熵模型从真实自然的语音流中自动识别汉语短语间停顿的方法。模型的特征集包含语音和词法两类特征,采用半自动的方式获得。首先由人工根据经验设计候选特征集,然后采用特征选择算法对候选特征进行筛选,选择更有效的特征构成最终特征集,并训练生成用于汉语短语间停顿识别的最大熵模型。3组实验的结果表明,模型能够取得比较满意的短语间停顿识别效果。     

基于最大熵模型的中文姓名识别

最大熵模型可以对非常广泛的自然语言现象建立概率模型，该文提出了一种使用统计模型的中文姓名识别方法，采用最大熵原则建立统计模型，并运用该模型计算姓名概率。系统采用真实语料进行开放测试的结果表明，在保证一定准确率的情况下，系统召回率可达90%以上。     

基于相对熵和二元熵的多粒度直觉语言TODIM决策方法

针对评估信息为多粒度直觉语言集的决策问题,提出一种基于相对熵和二元熵的TODIM方法。该方法首先定义了直觉语言数的相对熵和二元熵,以度量决策信息的差异和不确定性;其次,构建了基于相对熵和二元熵的专家赋权模型,并建立了主观权重完全已知、部分已知和完全未知场景下的属性赋权模型;最后,为集结多粒度群体决策信息,提出了多粒度直觉语言加权算术平均(MIL-WAA)算子。算例分析表明,该方法能够较好地度量决策信息的不确定性和差异性,并考虑了决策者的有限理性行为,具有一定的合理性和有效性。