基于累积Logistic 回归分析的文本段落聚类策略研究

提出一种新的文本段落聚类策略,该策略采用多特征融合思想尽可能多地挖掘段落内的特征,并采用累积Logistic 回归分析方法来拟合这些特征与段落相似度之间的内在关联,使得段落相似度计算的结果更为理想.最后采用层次聚合聚类算法中的complete-link方法对段落集合进行聚类处理.通过网络真实文本进行了段落相似度度量实验和段落聚类实验,实验结果显示了方法的可行性.     

一种基于蚁群聚类的多模板图像边缘融合方法

根据所采集图像的离散性特点,提出了一种基于蚁群聚类的新型的多模板图像边缘融合方法.该算法综合了图像聚类的方法和模板边缘提取的方法,运用蚁群聚类的思想来提取图像的边缘特征,并进一步融合多种边缘提取模板来设置聚类的启发式引导函数和初始聚类中心,从而避免了传统蚁群算法聚类搜索时的盲目性.系列仿真实验验证了这一方法的可行性和有效性.     

基于模式识别的自适应时钟重置检测算法

为解决网络测量中因测量主机的时钟重置导致单向时延测量误差的问题,根据时钟重置发生的不定时性和发生次数未知的特征,提出了一种基于模式识别的自适应时钟重置检测算法（PRBA）。该算法基于模式识别中的最大最小距离聚类算法对单向时延测量值进行聚类,同时利用时间序列技术中的一种低通滤波器过滤出噪声区间,从而有效识别网络行为特征,自适应地检测时钟值瞬时调整方式的时钟重置。实验结果表明,与现有同类算法相比,该算法具有较高准确性和自适应性。     

一种基于特征聚类和评价的轴承寿命预测新方法

在预测轴承寿命时,使提取的特征和剩余寿命保持高相关性,并使不同的特征之间保持低相关性,是有利于提升轴承寿命预测精度的。为解决单一的特征评价方法对后者考虑不足的问题,提出了一种基于相关性改进Kmeans聚类算法(correlation-based improved Kmeans cluster algorithm, Corr-Kmeans)和初始聚类中心确定方法,并与特征评价相结合,最终提出一种基于特征聚类和评价的轴承寿命预测新方法。首先利用卷积自编码对频域信息提取初始特征,用Corr-Kmeans对初始特征按相关性进行聚类,使得聚类后的特征类内相关性高,而类间相关性低;其次,使用相关性、单调性和鲁棒性3个指标来综合评价每一类中的特征,按照筛选阈值将得分较高的特征从每一类中分别选出,组成用于训练与预测的特征子集;最后采用LSTM(long short-term memory, LSTM)网络对轴承剩余寿命进行预测。在一个轴承加速寿命试验的公开数据集上使用留一法进行验证,利用对比试验证明了该方法在预测轴承剩余寿命上的有效性。     

空间相关噪声下信源个数的聚类检测算法

针对空间相关噪声情况,利用两个独立阵列之间噪声不相关的特性,采用联合协方差矩阵的规范相关系数作为聚类特征,提出了一种基于模糊c均值聚类的信源个数检测方法.并详细分析了应用Fuzzy-c-Means(FCM)聚类算法进行信源个数检测的3个问题: 聚类的趋势、有效性和聚类中心的初始化.与经典算法相比,本文算法有较好的角度分辨力和检测性能.仿真结果证明了该方法的有效性和鲁棒性.     

模糊中心聚类的模式识别学习方法

基于一个约束条件下的非线性规划问题的优化计算思想，把模糊中心聚类中计算输入矢量与中心的距离来实现聚类作为一种优化计算问题，证明了模糊中心聚类方法，取一个适当的属函数，其聚类中心vi为模糊聚类中心价值函数的极小值，推导出了基于模糊中心聚类的模式识别的无导师递推学习方法，提出了模糊中心聚类模式分类神经网络结构，该网络可以实现并行数据处理和模式分类的软划分和硬划分。     

基于知识迁移的深度学习无线电信号聚类方法

现有的无线电信号调制识别方法在先验数据不足时通常很难对无类标信号进行有效识别。针对这个问题,本文提出了一种基于知识迁移的深度学习无线电信号聚类方法(DTC)。该方法基于样本对比,分析样本间的相似性,并利用卷积神经网络(CNN)提取无线电信号的特征,同时设计了一种预训练框架,通过迁移同领域数据集的知识,有效提升了CNN特征提取能力,实现了引导聚类方向、提升聚类性能的目标。实验结果表明,该方法在多个公开数据集上的聚类性能都显著优于现有的聚类方法。与现有方法相比,DTC在RML 2016.10A和RML 2016.04C数据集上的聚类精度分别提升了30.34%和28.04%。     

一种中文文本聚类算法的研究

聚类是一种无指导的分类方法,在没有预先定义好分类的情况下,将一个大的数据集合分成若干个簇,要求数据在同一个簇中相似度尽可能大,而不同簇之间相似度尽可能小。聚类作为数据挖掘的一种重要方法,现在越来越被人们所重视。目前常见的聚类方法有:基于划分的聚类方法、基于层次的聚类方法、基于局部的聚类方法和基于模型的聚类方法,吸取各类聚类算法的实质,提出一种预设阀值,逐一归类的简单聚类实现算法,并在后端对聚类结果做精确行处理,经实验验证该方法能达到一定的聚类效果。     

基于纹理特征的钢丝绳图像分割方法

针对复杂背景下钢丝绳图像难以准确分割的问题,提出一种新的基于纹理特征的图像分割方法.首先,采用局部二进制模式(Local Binary Pattern,LBP)特征直方图的一阶熵、二阶熵作为LBP特征的统计测度,降低LBP特征的维数.同时选用边缘密度作为纹理描述的特征之一,弥补LBP算子提取纹理特征不足,抗干扰能力差的缺点.然后以上述纹理特征构成特征矢量,采用模糊C-均值(Fuzzy C-Mean,FCM聚类算法进行聚类分割.在实验中,对比了该算法与灰度共生矩阵、传统LBP算子在钢丝绳图像分割中的效果.结果表明,该算法可以有效地对钢丝绳图像进行纹理分割,并能取得良好的边界定位效果,性能优于另外两种算法.     

基于短语相似度的统计机器翻译模型

针对基于短语的统计机器翻译(SMT)模型中由于采用精确匹配策略导致的短语稀疏问题,提出了一种基于短语相似度的统计机器翻译模型.该模型将基于实例的翻译方法引入到统计机器翻译中.翻译时,对于训练语料库中未出现过的短语,通过计算源语言短语之间的相似度,采用模糊匹配策略从短语表中查找相似的实例短语,并根据实例短语为其构造翻译.与精确匹配策略相比,利用相似度进行模糊匹配增加了对短语表的利用程度,缓解了短语稀疏问题.实验表明,该模型能够明显地提高统计机器翻译的质量,效果超过了当前最好的短语系统"摩西(Moses)".     

A Hybrid Method of Coreference Resolution in Information Security

In the field of information security, a gap exists in the study of coreference resolution of entities. A hybrid method is proposed to solve the problem of coreference resolution in information security. The work consists of two parts: the first extracts all candidates (including noun phrases, pronouns, entities, and nested phrases) from a given document and classifies them; the second is coreference resolution of the selected candidates. In the first part, a method combining rules with a deep learning model (Dictionary BiLSTM-Attention-CRF, or DBAC) is proposed to extract all candidates in the text and classify them. In the DBAC model, the domain dictionary matching mechanism is introduced, and new features of words and their contexts are obtained according to the domain dictionary. In this way, full use can be made of the entities and entity-type information contained in the domain dictionary, which can help solve the recognition problem of both rare and long entities. In the second part, candidates are divided into pronoun candidates and noun phrase candidates according to the part of speech, and the coreference resolution of pronoun candidates is solved by making rules and coreference resolution of noun phrase candidates by machine learning. Finally, a dataset is created with which to evaluate our methods using information security data. The experimental results show that the proposed model exhibits better performance than the other baseline models.     

Predicting Simplified Thematic Progression Pattern for Discourse Analysis

The pattern of thematic progression, reflecting the semantic relationships between contextual two sentences, is an important subject in discourse analysis. We introduce a new corpus of Chinese news discourses annotated with thematic progression information and explore some computational methods to automatically extracting the discourse structural features of simplified thematic progression pattern (STPP) between contextual sentences in a text. Furthermore, these features are used in a hybrid approach to a major discourse analysis task, Chinese coreference resolution. This novel approach is built up via heuristic sieves and a machine learning method that comprehensively utilizes both the top-down STPP features and the bottom-up semantic features. Experimental results on the intersection of the CoNLL-2012 task shared dataset and the CDTC corpus demonstrate the effectiveness of our proposed approach.     

On Increasing the Resolution of Optical Receivers

Abstract

One approach for increasing the temporal resolution of optical receivers is discussed. It is based on a model of a photomultiplier with a set of resonators excited by secondary electrons together with retrieving algorithms. The resolution of such a system does not depend on the time-broadening of the secondary electron packets. In some cases, to increase the resolution it is not necessary to improve such parameters of the photomultiplier as time-broadening, jitter, etc., where the resolution of the photon detectors are defined with respect to single output video pulses.     

Shape-dependent regularization for the retrieval of atmospheric state parameter profiles

We present an adaptive regularization approach to retrieve vertical state parameter profiles from limb-sounding measurements with high accuracy. This is accomplished by introducing a dedicated regularization functional based on a reasonable assumption of the profile characteristics. The approach results in shape-dependent weighting during least-squares computations and relies on a Cholesky decomposition of a preselected L(T)L matrix. Our method is compared with established regularization functionals such as optimal estimation and Tikhonov with respect to errors and achievable height resolution. The results show an improved height resolution of the retrieved profiles together with a reduction of absolute and relative errors obtained by test-bed simulations.     

CCD高分辨成像的梯度解析法

基于序列图像重建的高分辨成像技术需要获取互有位移的序列图像,并需利用图像重建算法进行高分辨率图像重建。为此,设计了一种利用压电陶瓷体控制CCD位移来获取互有微小位移的序列图像的装置,在此基础上提出了一种基于梯度理论从序列图像重建高分辨率图像的算法——梯度解析法。该解析法根据图像灰度场梯度理论,同时考虑了图像的更高频成分,提高了重建图像分辨率。仿真实验显示,该算法与已有算法相比,重建图像的失真度降低40％以上。     

High-resolution digital holography utilized by the subpixel sampling method

A novel (to our knowledge) approach for resolution improvement in digital holography is presented in this paper. The proposed method is based on recording the incoming interference field on a complementary metal-oxide semiconductor (CMOS) camera with subpixel resolution. The method takes advantage of the small pixel size of the CMOS sensor, while overcoming the reduced fill factor. This paper describes the experimental and numerical procedures. The improvement of the obtainable optical resolution, image quality, and phase measurement accuracy are demonstrated within this paper.     

Linear approach to axial resolution in elasticity imaging

Thus far axial resolution in elasticity imaging has been addressed only empirically. No clear analytical approaches have emerged because the estimator is non-linear in the data, correlation functions are nonstationary, and system responses vary spatially. This paper describes a linear systems approach based on a small-strain impulse approximation that results in the derivation of a local impulse response (LIR) and local modulation transfer function (LMTF). Closed-form solutions for strain LIR are available to provide new insights on the role of instrumentation and processing on axial strain resolution. Novel phantom measurements are generated to validate results. We found that the correlation window determines axial resolution in most practical situations, but that the the same system properties that determine B-mode resolution ultimately limit elasticity imaging.     

In-situ measurements of the radioactive fallout deposit

An improved method to determine radionuclide concentrations in soil and the radioactive fallout deposit is presented. The approach is based on in-situ gamma-ray spectrometric measurements performed with a portable high resolution gamma spectrometer and on calculations of the depth distribution based on the energy dependence of the attenuation of gamma rays in soil. The results are compared with laboratory analysis of collected soil samples.     

A triangle mesh-based corner detection algorithm for catadioptric images

As with conventional images corner detection is an important aspect of many computer vision problems involving catadioptric images. However, classical image processing algorithms are no longer appropriate for catadioptric images due to nonuniform resolution and distortions of catadioptric images. In this paper, we propose a novel approach to corner detection for catadioptric images based on triangle mesh. First, we transform catadioptric images to spherical images by combining an improved projection model for central catadioptric cameras with triangle mesh for a unit sphere. Spherical images yield a spatially uniform resolution domain for processing catadioptric images. Then, based on the topology of a triangle mesh, variations of light intensity with respect to directions for each image patch are measured to detect corners. The proposed algorithm addresses problems of catadioptric image processing caused by non-uniform resolution and distortions of these images. Experimental results showed that comparing to widely used methods, the triangle mesh-based corner detection algorithm can achieve higher repeatability rate relative to different imaging condition changes.     

Ultrasound synthetic aperture imaging: monostatic approach

An ultrasound synthetic aperture imaging method based on a monostatic approach was studied experimentally. The proposed synthetic aperture method offers good dynamical resolution along with fast numerical reconstruction. In this study complex object data were recorded coherently in a two-dimensional hologram using a 3.5 MHz single transducer with a fairly wide-angle beam. Image reconstruction which applies the wavefront backward propagation method and the near-field curvature compensation was performed numerically in a microcomputer using the spatial frequency domain. This approach allows an efficient use of the FFT-algorithms. Because of the simple and fast scanning scheme and the efficient reconstruction algorithms the method can be made real-time. The image quality of the proposed method was studied by evaluating the spatial and dynamical resolution in a waterbath and in a typical tissue-mimicking phantom. The lateral as well as the range resolution (-6 dB) were approximately 1 mm in the depth range of 30-100 mm. The dynamical resolution could be improved considerably when the beam width was made narrower. Although it resulted in a slightly reduced spatial resolution this compromise has to be done for better resolution of low-contrast targets such as cysts. The study showed that cysts as small as 2 mm by diameter could be resolved