变精度粗糙集的加权KNN文本分类算法

针对训练样本较大时KNN算法(K最近邻算法)的分类效率较低和训练样本不均衡时算法的分类性能受到影响这两个问题,提出变精度粗糙集的加权KNN文本分类算法。利用变精度粗糙集上下近似的概念,将各个类别的训练集分为β正区域和β边界域。分类过程中根据测试样本与样本中心的相似度,得到样本的归属区域。其中属于β正区域的样本可以直接判断其类别,其它区域的样本用基于数量加权的KNN算法判断其类别。实验结果表明,该算法能有效提高分类的性能和效率。

改进的CK度量套件

为弥补CK套件中每类加权方法数量、继承树深度和方法内聚缺乏度等度量指标的缺陷,提出了如下改进的度量指标:每类加权方法参数复杂性(WMACC)、每类加权方法属性复杂性(WACC)、相对继承树深度(RDIT)、方法内聚性(CM)和封装率(ER).用Java的Util包中的类作为测试数据比较了ICK度量套件和CK套件,结果表明,WMACC和CM更准确地反映了方法的复杂性和内聚性,RDIT更适合于纯面向对象系统的环境.WACC和ER提供了对属性复杂性和封装性的度量.

Improving signal to noise ratio of seismic profiles using an image processing technique

This paper introduces an image processing technique into seismic data processing as a noise attenuation technology. The image separation of the seismic profile is obtained by using grating operators based on different time dips and a set of relative single dip profiles is obtained. A high signal to noise ratio profile can be obtained during reconstruction by statistical weighting. With further processing analysis and geological study, a high signal to noise profile that can meet geological requirements can be produced. The real data examples show that the signal to noise ratio of the profile is greatly improved, the resolution of the profile is maintained, and the fault terminations are much clearer after using the image processing method.

Comparison of Interpolation Methods for the Study of Forest Variables Using a Geographic Information System

Interpolation methods in ArcGIS_ESRI Geostatistical Analyst tool allow obtaining unknown values at unsampled points from observed data and generating continuous surfaces. In this paper, forest data variables as tree height and diameter measured in two plots in Central Mountains in Spain. These data were georeferenced to obtain maps that can visualize the spatial variability of these forest variables. In order to evaluate the best interpolation method that could adequately explain the spatial variability of those variables, two interpolation methods were studied： inverse results was made by means of statistical methods to analyze distance weighted （IDW） and Ordinary Kriging （OK）. A comparison of residuals. Results with the kriging method were slightly better.

Evaluating gravity gradient components based on a reweighted inversion method

In gravity gradient inversion,to choose an appropriate component combination is very important,that needs to understand the function of each component of gravity gradient in the inversion.In this paper,based on the previous research on the characteristics of gravity gradient components,we propose a reweighted inversion method to evaluate the influence of single gravity gradient component on the inversion resolution The proposed method only adopts the misfit function of the regularized equation and introduce a depth weighting function to overcome skin effect produced in gravity gradient inversion.A comparison between different inversion results was undertaken to verify the influence of the depth weighting function on the inversion result resolution.To avoid the premise of introducing prior information,we select the depth weighting function based on the sensitivity matrix.The inversion results using the single-prism model and the complex model show that the influence of different components on the resolution of inversion results is different in different directions,however,the inversion results based on two kind of models with adding different levels of random noise are basically consistent with the results of inversion without noises.Finally,the method was applied to real data from the Vinton salt dome,Louisiana,USA.

Augmented Kierlik-Rosinberg Fundamental Measure Functional and Extension of Fundamental Measure Functional to Inhomogeneous Non-hard Sphere Fluids

From point of view of weighted density procedure, it is guessed that a Percus-Yevick （PY） compressibility excess free energy density, appearing in the Kierlik Rosinberg type fundamental measure functional （KR-FMF） and expressed in terms of scaled particle variables, can be substituted by a corresponding expression dictated by a more accurate Mansoori Carnahan-Starling Leland （MCSL） equation of state, while retaining the original weighting functions; it is numerically indicated that the resultant undesirable non-self-consistency between the PY type weighting function and MCSL type excess free energy density had no bad effect on the performance of the resultant augmented KR-ffMF which, on the one hand, preserves the exact low-density limit of the original KR-FMF and holds a high degree of pressure self-consistency, on the other hand, improves significantly, as expected, the predictions of density profile of hard sphere fluid at single hard wall contact location and its vicinity, and of the bulk hard sphere second order direct correlation function （DCF）, obtained from functional differentiation. The FMF is made applicable to inhomogeneous non-hard sphere fluids by supplementing a functional perturbation expansion approximation truncated at the lowest order with summation of higher order terms beyond the lowest term calculated by the FMF for an effective hard sphere fluid; the resultant extended FMF only needs second order DCF and pressure of the fluid considered at coexistence state as inputs, consequently is applicable whether the considered temperature is above critical point or below critical point. The extended MCSL-augmented KR-FMF is found to be endowed with an excellent performance for predictions of density profile and surface tension by comparing the present predictions of these two quantities with available computer simulation data for inhomogeneous hard core attractive Yukawa fluid and Lennard-3ones fluid.

DISTRIBUTED QUANTIZED CONSENSUS FOR AGENTS ON DIRECTED NETWORKS

Communication bandwidth and network topology are two important factors that affect performance of distributed consensus in multi-agent systems.The available works about quantized average consensus assume that the adjacency matrices associated with the digraphs are doubly stochastic,which amounts to that the digital networks are balanced.However,this assumption may be unrealistic in practice.In this paper,without assuming double stochasticity,the authors revisit an existing quantized average consensus protocol with the logarithmic quantization scheme,and investigate the quantized consensus problem in general directed digital networks that are strongly connected but not necessarily balanced.The authors first derive an achievable upper bound of the quantization precision parameter to design suitable logarithmic quantizer,and this bound explicitly depends on network topology.Subsequently,by means of the matrix transformation and the Lyapunov techniques,the authors provide a testable condition under which the weighted average consensus can be achieved with the proposed quantized protocol.

Local Existence of Smooth Solutions to the FENE Dumbbell Model

The author proves the local existence of smooth solutions to the finite extensible nonlinear elasticity (FENE) dumbbell model of polymeric flows in some weighted spaces if the non-dimensional parameter b > 2.