Estimation on principal component of multi-collinearity Gauss-Markov model based on minimum description length

Gauss-Markov model is frequently used in data analysis; the analysis and estimation of its parameters is always a hot issue. Based on the information theory and from the viewpoint of optimal information on description—minimum description length, this paper discusses a case: where there is multi-collinearity in the coefficient matrix, principal component estimation is used to estimate and select the original parameters, so as to reduce its multi-collinearity and improve its credibility. From the viewpoint of minimum description length, this paper discusses the approach of selecting principal components and uses this approach to solve a practical problem.

Minimum Description Length Methods in Bayesian Model Selection: Some Applications

Computations involved in Bayesian approach to practical model selection problems are usually very difficult. Computational simplifications are sometimes possible, but are not generally applicable. There is a large literature available on a methodology based on information theory called Minimum Description Length (MDL). It is described here how many of these techniques are either directly Bayesian in nature, or are very good objective approximations to Bayesian solutions. First, connections between the Bayesian approach and MDL are theoretically explored;thereafter a few illustrations are provided to describe how MDL can give useful computational simplifications.

Gaussian mixture model clustering with completed likelihood minimum message length criterion

An improved Gaussian mixture model （GMM）- based clustering method is proposed for the difficult case where the true distribution of data is against the assumed GMM. First, an improved model selection criterion, the completed likelihood minimum message length criterion, is derived. It can measure both the goodness-of-fit of the candidate GMM to the data and the goodness-of-partition of the data. Secondly, by utilizing the proposed criterion as the clustering objective function, an improved expectation- maximization （EM） algorithm is developed, which can avoid poor local optimal solutions compared to the standard EM algorithm for estimating the model parameters. The experimental results demonstrate that the proposed method can rectify the over-fitting tendency of representative GMM-based clustering approaches and can robustly provide more accurate clustering results.

Supersonic Two-Dimensional Minimum Length Nozzle Design at High Temperature. Application for Air

When the stagnation temperature of a perfect gas increases, the specific heat ratio does not remain constant any more, and start to vary with this temperature. The gas remains perfect, its state equation remains always valid, except it will name in more calorically imperfect gas or gas at High Temperature. The goal of this work is to trace the profiles of the supersonic Minimum Length Nozzle with centered expansion when the stagnation temperature is taken into account, lower than the threshold of dissociation of the molecules and to have for each exit Mach number several nozzles shapes by changing the value of the temperature. The method of characteristics is used with a new form of the Prandtl Meyer function at high temperature. The resolution of the obtained equations is done by the second order of fmite differences method by using the predictor corrector algorithm. A study on the error given by the perfect gas model compared to our model is presented. The comparison is made with a calorically perfect gas for goal to give a limit of application of this model. The application is for the air.

Genetic Data Clustering Based on Minimum Coding Length

[Objective] This paper aimed to provide a new method for genetic data clustering by analyzing the clustering effect of genetic data clustering algorithm based on the minimum coding length. [Method] The genetic data clustering was regarded as high dimensional mixed data clustering. After preprocessing genetic data, the dimensions of the genetic data were reduced by principal component analysis, when genetic data presented Gaussian-like distribution. This distribution of genetic data could be clustered effectively through lossy data compression, which clustered the genes based on a simple clustering algorithm. This algorithm could achieve its best clustering result when the length of the codes of encoding clustered genes reached its minimum value. This algorithm and the traditional clustering algorithms were used to do the genetic data clustering of yeast and Arabidopsis, and the effectiveness of the algorithm was verified through genetic clustering internal evaluation and function evaluation. [Result] The clustering effect of the new algorithm in this study was superior to traditional clustering algorithms, and it also avoided the problems of subjective determination of clustering data and sensitiveness to initial clustering center. [Conclusion] This study provides a new clustering method for the genetic data clustering.

Flow Characteristics in Compact Thermal Spray Coating Systems with Minimum Length Nozzle

In this study, numerical analysis is performed to adopt the equivalence ratio on the high velocity oxygen fuel （HVOF） thermal spray coating systems equipped with a minimum length nozzle. The analysis is applied to investigate the axisymmetric, steady-state, turbulent, and chemically combusting flow both within the torch and in a free jet region between the torch and the substrate to be coated. The combustion is modeled using a single-step and eddy-dissipation model which assumes that the reaction rate is limited by the turbulent mixing rate of the fuel and oxidant. As the diameter of the nozzle throat is increased, the location of the Mach shock disc moves backward from the nozzle exit. As the throat diameter and the divergent portion are 6 mm and 8 mm, respectively, the pressure in the HVOF system is the lowest at the chamber and the expanding gas is steadily maintained with both high velocity and high temperature for different equivalence ratios. Thus, relatively minor amendments of the equivalence ratio and the geometry of HVOF can lead to improved control over coating characteristics.

基于双阵元天线的ADS-B解交织投影算法

随着广播式自动相关监视(Automatic Dependent Surveil lance-Broadcast,ADS-B)技术的普及,交通密集区域内多条ADS-B信号的交织问题难以避免。为了降低设备成本并提高系统在低成本条件下的可用性,针对双阵元天线三条ADS-B信号交织问题,提出了双阵元天线投影算法。首先,上述方法基于频域插值理论与最小描述长度准则进行信源数估计;然后,采用功率倒置算法进行空域滤波,在未知信号波达方向的条件下抑制非期望信号,从而将双天线三信号的解交织问题转化为单天线双信号的解交织问题;最后,由虚拟多通道法扩维后,基于改进的投影算法分离交织信号。仿真验证了以上算法的有效性。

基于多PRI脉压回波的单通道雷达干扰分离方法

研究了一种基于盲源分离的单通道雷达有源相干干扰抑制方法。针对伴随式干扰场景,在间歇采样转发干扰、部分脉冲密集转发干扰等有源相干干扰并存的条件下,建立了基于脉冲压缩的多脉冲重复间隔采样抗主瓣有源相干干扰模型并对其可行性进行了证明。针对现有单通道盲源分离方法在时域进行分离效果有限的问题,提出对脉压后的距离维数据进行盲源分离的思想,可在一定程度上抑制噪声对盲源分离的影响。同时,由于目标信号和干扰信号在脉冲压缩后重合度较低,可获得更好的分离效果。仿真和实测结果表明,在多个干扰源存在的条件下,本方法可有效分离真实目标回波和干扰回波。当信噪比大于5 dB时,目标脉压回波相似系数接近100%;选取虚拟通道数越多,分离效果越好。

基于最小描述长度原则的二阶段轨迹降维算法

针对现有轨迹简化方法不能统一不同长度轨迹样本维度的问题,本文提出了一种两次使用最小描述长度原则的轨迹简化算法,即基于最小描述长度原则的二阶段轨迹简化算法.第一阶段,基于最小描述长度原则的近似轨迹分割算法被使用,对原始单个轨迹样本进行第一次降维处理;第二阶段,再次使用最小描述长度原则迭代地删除原始轨迹上具备极少时空语义信息的位置坐标点,从而达到统一维度的目的.实验结果表明,该算法既能统一不同长度轨迹样本的维度,又能有效保留原始轨迹中的关键时空语义信息;并且当第一阶段降维算法对轨迹样本压缩程度适当的情况下,第二阶段算法仅会增加极少的时间成本.

考虑最小尺寸控制的压力驱动柔顺机构拓扑优化设计

拓扑优化设计的压力驱动柔顺机构拓扑构型容易出现类铰链结构,导致难以制造加工。为了满足制造工艺要求,提出一种考虑最小尺寸控制的压力驱动柔顺机构拓扑优化设计方法。采用改进的固体各向同性材料惩罚模型,利用达西定律结合排水项计算流体压力载荷,以机构的互应变能最大化和应变能最小化为优化目标,采用Otsu算法和拓扑细化算法提取柔顺机构的骨架特征,从而构建最小特征尺寸控制,以结构体积和最小特征尺寸作为约束,建立考虑最小尺寸控制的压力驱动柔顺机构拓扑优化模型,采用移动渐近线算法进行压力驱动柔顺机构拓扑优化问题求解。数值算例结果表明,所提设计方法获得的压力驱动柔顺机构最小特征尺寸满足约束,能够有效地抑制类铰链结构,并且分析不同最小控制尺寸对柔顺机构拓扑优化结果影响规律。

山区高速(特)长隧道及隧道群段换道方案探讨

为解决山区高速公路(特)长隧道及隧道群路段长时间跟车影响安全性、运行效率的发挥,同时为提高驾驶舒适性与驾乘体验,对云南省高速公路隧道数量和典型隧道可换道位置分布情况进行统计,重点针对已运营的保泸高速老营特长隧道、大丽高速花椒箐特长隧道及楚大高速九顶山特长隧道进行调研分析。根据交通流状况、线形指标条件、事故分布、驾驶需求等情况,结合实车驾驶试验观测与驾驶模拟仿真,开展隧道内可变道最小长度,隧道出入口段禁止变道白色实线长度及隧道群路段可变道路段布设等研究,得出隧道内允许换道的条件:特长隧道车道间,宜设置白色虚线允许变道;长隧道车道间,可设置白色虚线允许变道;中隧道车道间,宜设置白色实线禁止变道;短隧道车道间,应设置白色实线禁止变道;螺旋隧道,宜设置白色实线禁止变道。隧道内白色虚线的长度应不小于249 m,宜不小于399 m;2座隧道间距≥400 m时,隧道间可设置白色虚线允许变道。

GB/T5972--2016不同捻制类型钢丝绳报废断丝数间规律研究

对同类别编号下的交互捻钢丝绳、同向捻钢丝绳,在6d长度范围内或30d长度范围内,GB/T5972—2016显示,一定服役条件下钢丝绳达到报废程度的最少可见断丝数存在约2:1的固定比例关系。分析认为是因为当检查断丝钢丝绳长度为整数倍钢丝绳捻距时,在交互捻钢丝绳、同向捻钢丝绳单根股中二次螺旋钢丝出现的频次比也存在约2:1的固定比例关系。虽然标准给出同类别编号下的交互捻钢丝绳、同向捻钢丝绳达到报废程度的最少可见断丝数不同,但这并不意味着捻制类型会影响钢丝绳使用安全性。在1个钢丝绳捻距上,捻制类型对单根股中二次螺旋钢丝可见频次的贡献为±1,同向捻时为-1,交互捻时为+1。N与N±1有各自清晰含义。

山区狭长地形条件下回转绕行互通方案可行性探讨

对山区狭长地形条件下互通立交设计方案进行了探讨,得出辅助车道回转绕行互通方案与地形、地貌的协调性好,规模小,投资少,优势明显,可作为狭长地形条件下互通优选方案。对通行能力进行分析,得出交织区长度与流率比、交织区车流密度关系,以及不同行车速度、不同流率比下辅助车道最小长度,可以类似项目设计及研究提供依据。

一种高效的属性图聚类方法

图是描述现实世界各类复杂系统的一种普适模型,且许多实际应用中的图是大规模的.图的聚类是理解、分析和可视化大规模图的关键技术之一.现实世界的图往往包含丰富的属性信息,如何综合结构和属性信息进行属性图的聚类是一个新的挑战.大多数的现有方法或者将结构和属性转化为距离,基于传统方法进行聚类;或者只考虑某一方面聚类.文中结合信息论中最小长度原则,基于遗传算法,提出一种高效的属性图聚类方法GA-AGC.通过对属性图聚类问题建模,转化为最小描述长度原则问题;扩展标签传播方法作为遗传算法初始化方法,结合编码减小的局部变异方法,提出一种解决属性图聚类的遗传算法.文中方法无需设定聚类的数目,算法复杂度近似线性于结点和边的数目.真实数据集上的实验验证了算法的有效性和高效性.

基于三分状态MDL方法度量我国股市泡沫

通过改进的三分状态MDL(最小描述长度)方法,在马尔科夫模型下,对我国1991年4月3日～2008年9月23日的上证指数和深证成份指数的价格泡沫情况进行了分析,研究发现,略去考虑我国股市在建立初期的异常波动外,其泡沫主要集中于1996年3月～1997年5月,2000年以及2006年6月～2007年9月之间,且最后一个时段泡沫情况最为严重.

基于轨迹相似度的移动目标轨迹预测算法

传统的轨迹预测算法训练模型时需要耗费大量时间,且时空复杂度高、执行效率低,不能满足实时预测的需求。为此,提出一种改进的移动目标轨迹预测算法。基于欧氏距离进行轨迹相似度计算以提高预测准确性和实效性,根据最小描述长度原理对预测后的轨迹进行简化,优化运算效率。实验结果表明,该算法能准确预测移动目标的轨迹,并且具有较低的算法复杂度,适用于海量数据背景下的移动目标轨迹预测。

一种增量并行式动态图异常检测算法

图结构异常检测可以发现金融欺诈行为、网络入侵和可疑的社交行为。针对当前检测图异常算法的计算复杂度高、不能处理大规模动态图的缺点,研究并提出了一种增量并行式的算法以便更有效地发现和检测大规模动态图中的异常。该算法使用时间滑动窗口对图进行划分,在初始化阶段选取N个子图,使用最小描述长度(MDL)原理并行检测正常模式和异常模式,并行迭代地检测其他子图中的正常结构和异常结构。在多个大规模图数据集上的实验结果表明,检测动态图结构异常准确率达到96%,召回率达到85%,运行时间减少了一个数量级。同时还讨论了滑动窗口大小和并行数量对算法运行时间的影响。

基于MDL原理与混合遗传算法的Bayesian网络结构学习

从大型数据库中学习Bayesian网络结构是Bayesian网络应用的难点之一。在分析标准遗传算法与爬山算法各自优点与不足的基础上,将这两种算法相结合,以最小描述长度为评价函数,得到一种混合遗传算法,实现了它们的优势互补。文章给出了混合遗传算法的计算步骤,并通过对ALARM数据库学习得到的Bayesian网络结构。

基于最长公共子序列距离的主旨模式挖掘算法

针对现有主旨模式挖掘算法易受噪声干扰的问题,提出一种基于最长公共子序列距离的挖掘算法。在搜索过程中,该算法采用基于子序列距离判别的策略进行了有效的剪枝,对于非等长的候选模式,使用最小描述长度原则求其相关权重,据此选择出现频率最高、最能体现原时间序列特征的主旨模式。实验结果表明,与朴素式搜索相比,该算法的速度至少提升60%。

融合先验信息的贝叶斯网络结构学习方法

在贝叶斯网络结构学习的过程中,如何采集先验信息并合理利用它对于构建准确的网络结构非常重要。鉴于此,依据有先验信息的贝叶斯网络结构学习的三个环节:先验信息的采集、先验信息的融合和网络结构的优化,首先讨论了现有先验信息获取方法的不足,并提出了基于信念图的先验信息获取方法;其次针对所获取的先验信息通常具有一定的不确性,对最小描述长度测度进行了改进以融合非确定性先验信息;最后依据问题特性对模拟退火算法进行了适当的修改以更好地优化网络结构。实验表明,提出的结构学习方法能够有效地提高网络结构的学习精度。