基于类标号扩展的半监督特征选择算法

特征选择是数据挖掘、机器学习等领域的重要内容,在缺乏已标记样本的情况下,如何有效选择特征是一个非常值得研究的问题。基于集合间相关度与自相关度的定义,提出了一种新颖的半监督特征选择方法,从原始、少量、且已标记的训练样本出发,通过扩展类标号得到最终的聚类效果,采用复合的评价方法作为衡量特征子集的标准。大量实验结果表明,该算法是有效的。

并行设计中耦合活动集的割裂规划

为了提高耦合活动集割裂规划的有效性,针对现有割裂方法只考虑设计活动之间行为上依赖关系的局限性,提出基于互相关度和自相关度综合指标的割裂准则.在构造该准则的过程中,用三角模糊数的层次分析法(AHP)分析设计资源之间的关系;用互相关度指标分析设计活动与设计资源之间的关系;用自相关度指标分析设计活动之间的关系.最后,结合某型起重机改进设计的实例,对新的割裂准则进行说明和验证.实践表明,该方法能有效保证并行设计过程在资源约束条件下顺利进行,并为设计资源快速调度规划提供依据.

Study on the scales of heterogeneous geologic bodies in random media

In order to study the scale characteristics of heterogeneities in complex media, a random medium is constructed using a statistical method and by changing model parameters （autocorrelation lengths a and b）, the scales of heterogeneous geologic bodies in the horizontal and the vertical Cartesian directions may be varied in the medium. The autocorrelation lengths a and b represent the mean scale of heterogeneous geologic bodies in the horizontal and vertical Cartesian directions in the randQm medium, respectively. Based on this model, the relationship between model autocorrelation lengths and heterogeneous geologic body scales is studied by horizontal velocity variation and standard deviation. The horizontal velocity variation research shows that velocities are in random perturbation. The heterogeneous geologic body scale increases with increasing autocorrelation length. The recursion equation for the relationship between autocorrelation lengths and heterogeneous geologic body scales is determined from the velocity standard deviation research and the actual heterogeneous geologic body scale magnitude can be estimated by the equation.

基于双IRS辅助的毫米波MIMO通信信道估计

考虑到智能反射面(Intelligent Reflecting Surface,IRS)可以辅助毫米波通信,为提高信道估计的准确度,引入基于双智能反射面辅助的毫米波单用户多输入多输出(Multiple⁃Input Multiple⁃Output,MIMO)无线通信新场景。因为毫米波信道具有稀疏性,首先将原始毫米波窄带信道模型通过信道编码转换为压缩感知信道估计模型,并以感知矩阵自相关度为理论基础分析得出双IRS通信场景会提高信道估计性能的观点。通过使用基于匹配追踪的广义近似消息传递(Generalized Approximate Message Passing,GAMP)压缩感知算法降低了对信道先验信息的要求,从而降低信道估计的计算复杂度。仿真结果表明,双IRS信道估计性能显著优于单IRS,验证了理论推导的正确性。

Evaluation of Phonon-level Density of UO2 by Molecular Dynamics Simulation

Molecular dynamics calculation of UO2 in a wide temperature range are presented and dis- cussed. The calculated lattice parameters, mean square displacements, and dynamic prop- erty of phonon-level density of the velocity auto-correlation functions for UO2 are provided. The Morelon potential and the Basak potential are employed. It confirms that the cal- culated lattice parameters using the Basak potential are in nearly perfect agreement with the reported values. The models successfully predict mean square displacement and Bredig transition. Furthermore, the phonon-level density of uranium dioxide are discussed. The intensity of phonon-level density increases with temperature, and the properties of UO2 are characterized by large thermal vibrations rather than extensive disorder.

Modeling of rough surfaces with given roughness parameters

Modeling of rough surfaces with given roughness parameters is studied,where surfaces with Gaussian height distribution and exponential auto-correlation function(ACF) are concerned.A large number of micro topography samples are randomly generated first using the rough surface simulation method with FFT.Then roughness parameters of the simulated roughness profiles are calculated according to parameter definition,and the relationship between roughness parameters and statistical distribution parameters is investigated.The effects of high-pass filtering with different cut-off lengths on the relationship are analyzed.Subsequently,computing formulae of roughness parameters based on standard deviation and correlation length are constructed with mathematical regression method.The constructed formulae are tested with measured data of actual topographies,and the influences of auto-correlation variations at different lag lengths on the change of roughness parameter are discussed.The constructed computing formulae provide an approach to active modeling of rough surfaces with given roughness parameters.

Spatial Autocorrelation and Localization of Urban Development

A nonlinear analysis of urban evolution is made by using of spatial autocorrelation theory. A first-order nonlinear autoregression model based on Clark’s negative exponential model is proposed to show urban population density. The new method and model are applied to Hangzhou City, China, as an example. The average distance of population activities, the auto-correlation coefficient of urban population density, and the auto-regressive function values all show trends of gradual increase from 1964 to 2000, but there always is a sharp first-order cutoff in the partial auto- correlations. These results indicate that urban development is a process of localization. The discovery of urban locality is significant to improve the cellular-automata-based urban simulation of modeling spatial complexity.

Computation of Average Run Length for Residual-Based T^2 Control Chart for Multivariate Autocorrelated Processes

The expression of residual is obtained according to its dynamic response to mean shift, then the distribu- tion of T2 statistic applied to the residual is derived, thus the probability of the 7a statistic lying outside the control limit is calculated. The above-mentioned results are substituted into the infinite definition expression of the average run length （ARL）, and then the final finite ARL expression is obtained. An example is used to demonstrate the procedures of the proposed method. In the comparative study, eight autocorrelated processes and four different mean shifts are performed, and the ARL values of the proposed method are compared with those obtained by simulation method with 50 000 replications. The accuracy of the proposed method can be illustrated through the comparative results.

IDFT Numerical Simulation Method for Gaussian Rough Surface with Relatively Large Correlation Length

A numerical simulation method based on inverse discrete Fourier transform(IDFT)is presented for generating Gaussian rough surface with a desired autocorrelation function(ACF). The probability density function of the height distribution of the generated Gaussian surface and the root-mean-square height of the rough surface are also considered. It is found that the height distribution of the generated surface follows the Gaussian distribution, the deviation of the root-mean-square height of the modeled rough surface from the desired value is smaller than that of Patir's method, and the autocorrelation function of the modeled surface is also in good agreement with the desired autocorrelation function. Compared with Patir's method, the modeled surface generated by the IDFT method is in better agreement with the desired autocorrelation function, especially when the correlation length is relatively large.

ADS-B Anomaly Data Detection Model Based on Deep Learning and Difference of Gaussian Approach

Due to the influence of terrain structure,meteorological conditions and various factors,there are anomalous data in automatic dependent surveillance-broadcast(ADS-B)message.The ADS-B equipment can be used for positioning of general aviation aircraft.Aim to acquire the accurate position information of aircraft and detect anomaly data,the ADS-B anomaly data detection model based on deep learning and difference of Gaussian(DoG)approach is proposed.First,according to the characteristic of ADS-B data,the ADS-B position data are transformed into the coordinate system.And the origin of the coordinate system is set up as the take-off point.Then,based on the kinematic principle,the ADS-B anomaly data can be removed.Moreover,the details of the ADS-B position data can be got by the DoG approach.Finally,the long short-term memory(LSTM)neural network is used to optimize the recurrent neural network(RNN)with severe gradient reduction for processing ADS-B data.The position data of ADS-B are reconstructed by the sequence to sequence(seq2seq)model which is composed of LSTM neural network,and the reconstruction error is used to detect the anomalous data.Based on the real flight data of general aviation aircraft,the simulation results show that the anomaly data can be detected effectively by the proposed method of reconstructing ADS-B data with the seq2seq model,and its running time is reduced.Compared with the RNN,the accuracy of anomaly detection is increased by 2.7%.The performance of the proposed model is better than that of the traditional anomaly detection models.

窄带傅里叶变换光谱仪中平稳高斯噪声的理论分析

在本文所研究的空间调制型傅里叶变换光谱仪中,在窄的光谱带宽内进行探测可以有效提高光谱的分辨率.为了研究光源辐射噪声对系统的影响,本文将辐射噪声归纳为一种高斯窄带的平稳随机过程,利用线性系统分析方法,根据统计学原理推导了辐射噪声作为随机变量通过干涉系统前后的输入信噪比与输出信噪比.然后,对于一个具有窄带矩形光谱的系统进行了计算仿真,得出了信噪比增益随着光程差在不同自相关度和互相关度取值时的变化情况.仿真结果表明,不同光程差下的信噪比增益在相关度空间是一单调的平滑曲面,增益极值始终沿着相关度圆的半径和周线移动,并且经过一个光程差又回到初始的位置.根据对信噪比增益的分析,可以将噪声的相关度取值控制在某一范围之内,并作为系统光源设计与测试的依据.

Spatio-temporal evolution of population and urbanization in the countries along the Belt and Road 1950–2050

This paper uses data for the period 1950-2050 compiled by the United Nations Population Division together with methods including spatial autocorrelation analysis, hie- rarchical cluster analysis and the standard deviational ellipse, to analyze the spatio-temporal evolution of population and urbanization in the 75 countries located along the routes of the Silk Road Economic Belt and the 21st-century Maritime Silk Road, to identify future popula- tion growth and urbanization hotspots. The results reveal the following： First, in 2015, the majority of Belt and Road countries in Europe, South Asia and Southeast Asia had high population densities, whereas most countries in Central Asia, North Africa and West Asia, as well as Russia and Mongolia, had low population densities; the majority of countries in South Asia, Southeast Asia, Central Asia, West Asia and North Africa had rapid population growth, whereas many countries in Europe had negative population growth; and five Belt and Road countries are in the initial stage of urbanization, 44 countries are in the acceleration stage of urbanization, and 26 are in the terminal stage of urbanization. Second, in the century from 1950 to 2050, the mean center of the study area＇s population is consistently located in the border region between India and China. Prior to 2000, the trajectory of the mean center was from northwest to southeast, but from 2000 it is on a southward trajectory, as the population of the study area becomes more concentrated. Future population growth hotspots are predicted to be in South Asia, West Asia and Southeast Asia, and hotspot countries for the period 2015-2030 include India, China, Pakistan and Indonesia, though China will move into nega- tive population growth after 2030. Third, the overall urban population of Belt and Road coun- tries increased from 22% in 1950 to 49% in 2015, and it is expected to gradually catch up with the world average, reaching 64% in 2050. The different levels of urbanization in different countries display significant spatial dependency, and in the hundred-year period under con-sideration, this dependency increases before eventually weakening. Fourth, between 2015 and 2030, urban population hotspots will include Thailand, China, Laos and Albania, while Kuwait, Cyprus, Qatar and Estonia will be urban ＂coldspots.＂ Fifth, there were 293 cities with populations over 1 million located along the Belt and Road in 2015, but that number Js ex- pected to increase to 377 by 2030. Of those, 43 will be in China, with many of the others located in India, Indonesia and the eastern Mediterranean.

The Fulde-Ferrell-Larkin-Ovchinnikov states in s and d-wave superconductor with spin-dependent bandwidth imbalance on square lattice

We study numerically the phase diagram for s and d-wave fermionic superftuidity/superconductivity with spin-dependent band- width imbalance on a two-dimensional square-lattice. We also investigate the spontaneous space symmetry breaking states at low temperatures by solving the Bogoliubov-de Gennes equations. It is found that, the spatial configuration of the order parameter, both the uni-directional Fulde-Ferrell-Larkin-Ovchinnikov （FFLO） states and the two-dimensional FFLO state may show up in the presence of finite spin-dependent bandwidth imbalance. Moreover, we calculate the spectra of local density of states, and the experimental proposals of observing such FFLO states are therefore suggested.

Effects of Ionic Dependence of DNA Persistence Length on the DNA Condensation at Room Temperature

DNA persistence length is a key parameter for quantitative interpretation of the conformational properties of DNA and related to the bending rigidity of DNA.A series of experiments pointed out that,in the DNA condensation process by multivalent cations,the condensed DNA takes elongated coil or compact globule states and the population of the compact globule states increases with an increase in ionic concentration.At the same time,single molecule experiments carried out in solution with multivalent cations(such as spermidine,spermine)indicated that DNA persistence length strongly depends on the ionic concentration.In order to revolve the effects of ionic concentration dependence of persistence length on DNA condensation,a model including the ionic concentration dependence of persistence length and strong correlation of multivalent cation on DNA is provided.The autocorrelation function of the tangent vectors is found as an effective way to detect the ionic concentration dependence of toroidal conformations.With an increase in ion concentration,the first periodic oscillation contained in the autocorrelation function shifts,the number of segment contained in the first periodic oscillation decreases gradually.According to the experiments,the average long-axis length is defined to estimate the ionic concentration dependence of condensation process further.The relation between long-axis length and ionic concentration matches the experimental results qualitatively.