Model-Free Feature Screening via Maximal Information Coefficient (MIC) for Ultrahigh-Dimensional Multiclass Classification

It is common for datasets to contain both categorical and continuous variables. However, many feature screening methods designed for high-dimensional classification assume that the variables are continuous. This limits the applicability of existing methods in handling this complex scenario. To address this issue, we propose a model-free feature screening approach for ultra-high-dimensional multi-classification that can handle both categorical and continuous variables. Our proposed feature screening method utilizes the Maximal Information Coefficient to assess the predictive power of the variables. By satisfying certain regularity conditions, we have proven that our screening procedure possesses the sure screening property and ranking consistency properties. To validate the effectiveness of our approach, we conduct simulation studies and provide real data analysis examples to demonstrate its performance in finite samples. In summary, our proposed method offers a solution for effectively screening features in ultra-high-dimensional datasets with a mixture of categorical and continuous covariates.

基于一种距离相关的超高维生存数据Model-Free特征筛选

随着大数据时代的来临,数据维度爆炸式增长,超高维数据的降维问题逐渐成为众多研究领域的热点话题。由于响应变量通常存在右删失,处理超高维完全数据的降维方法在右删失数据中将不再适用。本研究提出一种新的基于距离相关能有效处理超高维右删失数据的特征筛选方法。首先利用距离相关系数计算每个协变量对响应变量的边际效应,建立与该系数有关的筛选指标,然后再根据事先确立的筛选准则进行特征筛选。提出的特征筛选方法不依赖任何模型结构假定,因此可以有效避免模型指定错误带来的不良后果。此外,该方法采用的距离协方差估计量是总体距离协方差的一个无偏估计,统计准确性和计算精度高。模拟和实证研究表明,提出的方法能在保留所有重要变量的前提下快速剔除与响应变量相关程度较弱的协变量,从而达到降低参数维数的目的。

Model-Free Ultra-High-Dimensional Feature Screening for Multi-Classified Response Data Based on Weighted Jensen-Shannon Divergence

In ultra-high-dimensional data, it is common for the response variable to be multi-classified. Therefore, this paper proposes a model-free screening method for variables whose response variable is multi-classified from the point of view of introducing Jensen-Shannon divergence to measure the importance of covariates. The idea of the method is to calculate the Jensen-Shannon divergence between the conditional probability distribution of the covariates on a given response variable and the unconditional probability distribution of the covariates, and then use the probabilities of the response variables as weights to calculate the weighted Jensen-Shannon divergence, where a larger weighted Jensen-Shannon divergence means that the covariates are more important. Additionally, we also investigated an adapted version of the method, which is to measure the relationship between the covariates and the response variable using the weighted Jensen-Shannon divergence adjusted by the logarithmic factor of the number of categories when the number of categories in each covariate varies. Then, through both theoretical and simulation experiments, it was demonstrated that the proposed methods have sure screening and ranking consistency properties. Finally, the results from simulation and real-dataset experiments show that in feature screening, the proposed methods investigated are robust in performance and faster in computational speed compared with an existing method.

Three-dimensional numerical simulation of crown spike due to coupling effect between bubbles and free surface

The motion of gas bubbles beneath a free surface will lead to a spike of fluid on the free surface. The distance of the bubbles to the free surface is the key factor to different phenomena. When the inception distance varies in some range, crown phenomenon would happen after the impact of weak buoyancy bubbles, so this kind of spike is defined as crown spike in the present paper. Based on potential flow theory, a three-dimensional numerical model is established to simulate the motion of the free-surface spike generated by one bubble or a horizontal line of two in-phase bubbles. After the downward jet formed near the end of the collapse phase, the simulation of the free surface is performed to study the crown spike without regard to the toroidal bubble＇s effect. Calculations about the interaction between one bubble and free surface agree well with the experimental results conducted with a high-speed camera, and relative error is within 15%. Crown spike in both single- and two-bubble cases are simulated numerically. Different features and laws of the motion of crown spike, depending on the bubble-boundary distances and the inter-bubble distances, have been investigated.

Isolation and Identification of Cancer Stem Cells from Human Osteosarcom by Serum-free Three-dimensional Culture Combined with Anticancer Drugs

The cancer stem cells(CSCs)from human osteosarcoma by serum-free three-dimensional culture combined with anticancer drugs were isolated and identified.The primary cells derived from human osteosarcoma were digested by trypsin to prepare a single-cell suspension,and mixed homogeneously into 1.2% alginate gel.Single-cell alginate gel was cultured with serum-free DMEM/F12 medium.Epirubicin(0.8μg/mL)was added to the medium to enrich CSCs.After cultured conventionally for 7 to 10 days,most of cells suspended in ...

Free Entropy Dimensions of Some von Neumann Algebras

In this paper, we estimate the free entropy dimension of the group yon Neumann algebra L（Zt）, which is less than 1/t,2 ≤t ≤ ＋∞. This data is identical with the free dimension defined by Dykema.

ANGULAR SPECTRUM PROPERTY OF SPONTANEOUS EMISSION IN A TWO-DIMENSIONAL UNDULATOR FREE-ELECTRON LASER

The angular spectrum of spontaneous emission in a two-dimensional undulator free-electron laser is analyzed theoretically. Numerical calculation shows that the 3-th harmonic spontaneous emission power density can be greatly enhanced by using a two-dimensional undulator, for which l=s, so the harmonic number can be selected by selecting l. Therefore, the higher harmonic operation of a free-electron laser can be realized selectively.

船舶有航速时三维时域格林函数计算方法研究

船舶工业CAE软件势流求解器的开发对势流计算方法的可靠性提出要求,本文针对船舶有航速时深水三维时域格林函数计算方法开展研究。首先引入三维时域自由面格林函数,其瞬时项的求解采用Hess&Smith方法,自由面记忆项的求解采用美国密歇根大学Beck团队的方法。其次,分区域推导出三维时域自由面格林函数记忆项及其导数的编程所用的数学展开表达式。然后,采用间接法模型求解分布源强和速度势,并采用压力直接积分法求出时域辐射力。最后,针对Wigley I船型开展了辐射力计算,并与公开的试验及计算结果进行对比,验证本文方法的可靠性。

Model-Free Feature Screening Based on Gini Impurity for Ultrahigh-Dimensional Multiclass Classification

It is quite common that both categorical and continuous covariates appear in the data. But, most feature screening methods for ultrahigh-dimensional classification assume the covariates are continuous. And applicable feature screening method is very limited;to handle this non-trivial situation, we propose a model-free feature screening for ultrahigh-dimensional multi-classification with both categorical and continuous covariates. The proposed feature screening method will be based on Gini impurity to evaluate the prediction power of covariates. Under certain regularity conditions, it is proved that the proposed screening procedure possesses the sure screening property and ranking consistency properties. We demonstrate the finite sample performance of the proposed procedure by simulation studies and illustrate using real data analysis.

Three-Dimensional Exact Solution for Dynamic Damped Response of Plates with Two Free Edges

A three-dimensional state space method has been developed for the calculation of dynamic response of plates with two free edges and two simply supported edges.A complex damping model was introduced, then the exact solutions which satisfy all the governing equations and boundary conditions were obtained.In order to overcome the difficulty of satisfying all the stress conditions at free edges, the displacement functions of free edges were assumed.The boundary conditions were strictly satisfied when the convergence rate was good.The computing time was evidently less than that of finite element method.The comparison of the solution with those of finite element method show that there is an excellent agreement for displacements.When the imaginary parts of normal stress deviated, the finite element results showed existence of shear stresses at top and bottom surfaces, and the boundary conditions of FEM model were not strictly satisfied.

Leading-edge receptivity of boundary layer to three-dimensional free-stream turbulence

The laminar-turbulent transition has always been a hot topic of fluid mechanics. Receptivity is the initial stage and plays a crucial role in the entire transition process. The previous studies of receptivity focus on external disturbances such as sound waves and vortices in the free stream, whereas those on the leading-edge receptivity to the three-dimensional free-stream turbulence (FST), which is more general in the nature, are rarely reported. In consideration of this, this work is devoted to investigating the receptivity process of three-dimensional Tollmien-Schlichting (T-S) wave packets excited by the three-dimensional FST in a flat-plate boundary layer numerically. The relations between the leading-edge receptivity and the turbulence intensity are established, and the influence of the FST directions on the propagation directions and group velocities of the excited T-S wave packets is studied. Moreover, the leading-edge receptivity to the anisotropic FST is also studied. This parametric investigation can contribute to the prediction of laminar-turbulent transition.

基岩中入射波三维时域反演及自由场构建

基岩入射波一维和二维时域反演不能反映入射方位角和斜入射角的空间任意性,导致构建的自由场可能与实际偏离较远,因此有必要开展基岩入射波三维时域反演研究。基于波函数组合法,利用基岩中P波、SV波和SH波的入射波时程表达地表控制点自由场,建立基于设计地震动的基岩入射波三维时域反演方法,并进一步构建了多波组合斜入射下的空间非一致自由场,分析了基于一维和二维反演方法构建的自由场偏差,揭示了不同反演和自由场构建方法对于沥青混凝土心墙土石坝地震响应的影响机制。结果表明:基于地震动三维反演方法构建的自由场具有空间非一致性,能够更全面地反映场地的地震动场情况。当入射方位角在0°~60°、斜入射角在40°~90°范围内,与三维反演方法相比,一维反演方法构建的特征点x向自由场偏差较大,位移峰值最大偏差为55.8%;当入射方向与某个水平坐标轴平行时,二维反演方法构建的特征点另一个水平向自由场偏差较大,位移峰值最大偏差为100%。介质泊松比增大,一维反演下x向自由场偏差减小,二维反演下x向自由场偏差增大,特征点位移峰值最大偏差分别为46%和36%。与基于地震动三维反演方法获得的心墙拉应力最大值相比,二维和一维反演下拉应力最大值分别减小了83.3%和20.0%。

基于无人机PPK技术在免像控条件下山区光伏地形图的生产

针对光伏场区高差大、部分区域植被覆盖、通信信号弱的特点,本文提出了一种基于动态后处理技术(PPK)在免像控条件下航测光伏地形图生产的方法。该方法首先在免像控的条件下对航测获取的无人机影像定位测姿系统(POS)数据和点云数据进行PPK解算;然后利用点云后处理、影像空中三角测量、数字微分纠正等技术生成高精度数字高程模型(DEM)和数字正射影像(DOM)成果;最后利用DEM和DOM成果构建三维模型,实现地形图生产。利用两个典型应用实例,验证了该方法可完全满足光伏地形图的精度要求。

三维自由弯曲成形管材损伤形式及成形质量对比试验研究

针对三维自由弯曲成形过程中成形装备的模具材料与管材材料匹配性差、管材弯曲成形时表面损伤严重的问题,为了提高管材成形质量,减轻表面损伤程度,研究了不同摩擦条件下管材损伤形式。对不同内衬材料与管材的摩擦副进行了摩擦试验,并基于试验结果对金属试样进行了表面形貌观察分析,获得了不同内衬材料对金属试样造成的损伤机制,分析得出了不同内衬材料与不同管材之间的最优匹配组,完成了三维自由弯曲成形装备模具内衬材料的优化,并进行了管材自由弯曲成形对比试验研究。结果表明,铝合金管材在钨钢内衬模具成形后、不锈钢管材在氧化锆内衬模具成形后的表面光洁无划伤,表面成形质量得到有效改善,同时发现适当减小偏心距、增大成形推进速度可以进一步改善管材表面成形质量。

基于FreeForm与3ds Max建模的三维打印

针对三维打印机对产品设计与制作周期的要求较高,传统的专业建模软件对于普通大众来说可实现性较差,不利于推广三维打印技术的问题,对比了传统建模软件与Free Form触觉建模系统的建模特点,提出了一种将Free Form系统和3ds Max软件相结合的建模方式。首先,在Free Form系统中设计具有自由曲面特性的基本黏土模型,并将设计完的模型按比例缩减10%后以STL格式导出;其次,将模型导入到3ds Max软件中,利用复合建模方法继续对模型进行外观设计和尺寸处理;最后,将模型重新加载到Free Form系统中进行模型修复和细节加工,输出经过近似处理的STL格式的三维模型。实验结果表明,该建模方法相对于传统专业建模软件下的建模,提高了建模速度,缩短了产品制作周期,建立的模型经Maker Bot Desktop软件切片处理后,可成功地转化为x3g格式的打印模型,并能在三维打印机上正确打印。

无碳小车的多维实践教学平台构建与实施

以全国大学生工程训练综合能力竞赛赛题“无碳小车”项目为载体,构建了无碳小车的多维实践教学平台。该实践教学平台包含“材料选择和应用”“工艺路线与加工精度”和“无碳小车设计和制作”3个实践教学模块,并对接多个课程单元,使学生所学的专业课程理论知识在实践中得到验证和强化。教学实施中,系统性整合课程内容,并通过“课堂—云课堂—实践”三元驱动形成合力,在教学过程中不断融入思政教育,实现学生的专业知识、实践能力、综合素质“三位一体”融合发展。

基于一维卷积神经网络的粉煤灰混凝土氯离子质量分数预测

为研究深度学习方法在氯离子质量分数预测中的应用,基于自然潮差环境下粉煤灰混凝土长期暴露试验获取了3150组自由氯离子质量分数数据,建立一维卷积神经网络(One-dimensional convolutional neural network,1D-CNN)模型用于预测粉煤灰混凝土氯离子质量分数。该模型分析了核函数和卷积层对1D-CNN预测精度的影响,研究了水灰比、暴露时间、粉煤灰掺量和渗透深度4个输入参数对粉煤灰混凝土自由氯离子质量分数预测结果的影响。实验结果表明:采用12个3×1卷积核及两层卷积层构建1D-CNN模型时,自由氯离子质量分数的预测结果最优;同时,应用最优的1D-CNN模型开展基于未测参数的自由氯离子质量分数预测,预测结果较为准确。因此,1D-CNN模型具有精度高和适用范围广泛的特点,能够为氯盐环境下混凝土中自由氯离子质量分数预测提供新的方法。

數字化導板引導游離端牙種植的精準度分析

目的評估使用數字化導板引導實施牙列遊離端種植的精準度。方法本研究收集2021年12月~2022年12月期間在澳門鏡湖醫院口腔科進行的數字化導板引導的游離端種植牙手術患者16例(植體16枚)。收集同期實施的非游離端種植牙手術患者16例(植體16枚)。使用Blue Sky Plan將患者術前設計數據與術後種植體掃描數據進行交叉比對,測量術前十設計位置與術後實際位置的差異,統計學分析游離端與非游離端植入的植體精準度差異。結果與術前設計植入位置比較,在牙列游離端植入種植體與非游離端植入種植體,其種植精準度相似,無統計學差異(P>0.05)。結論即使在牙列游離端,數字化導板都可以指導手術實現精確植入種植體,值得臨床推廣應用。

基于部分参数约束的粒子加速器隧道控制网平差

为控制隧道控制网平差时的误差累积问题,提高隧道控制网平差精度,设计了3种约束方案。首先,考虑到测站铅垂线之间的不平行性,依据控制网规模计算测站铅垂线与储存环中心铅垂线之间的夹角,在平差时顾及该约束,使平差模型更加严密。然后,为提高隧道控制网的平面精度,使用全站仪对距离较远的控制点进行长边夹角测量,在平差过程中附加长边夹角约束。其次,按照一等水准规范,对隧道地面点进行水准测量,基于水准测量结果,采用最小二乘转换方法,将控制网转换至水平面,进一步提高控制网竖向精度。最后,对周长为1360 m的隧道控制网进行仿真试验。结果表明,顾及测站垂差及水准点高程或高差平差值可提高控制网竖向精度;附加长边夹角约束可提高控制网平面精度;同时顾及测站垂差、长边夹角观测值及水准高程或高差平差值可提高控制点的点位精度,验证了约束条件的有效性。研究成果可为相关工程布设控制网提供参考。

改进无模型自适应迭代学习的直线电机滑模控制

二维直线电机具有非线性、多变量及强耦合等特点,其精确模型无法获得,同时,在实际过程中由于参数摄动及扰动而造成的运行不稳定现象,致使对其跟踪控制十分困难。针对以上问题,根据无模型自适应迭代学习控制不依赖被控系统精确数学模型的特点,及滑模控制具有可设计且与对象参数和扰动无关的滑动模态,提出一种改进无模型自适应迭代学习的滑模结构复合控制策略。在无模型自适应迭代学习控制方案的准则函数中加入误差变化率,并对其收敛性进行分析论证;然后在紧格式动态线性化基础上,设计指数趋近律滑模控制,使改进无模型自适应迭代学习的滑模控制复合策略能够克服不稳定现象并具有很强的鲁棒性,从而进一步提高系统响应速度和控制精度。最后通过仿真和实物验证,控制精度稳定在1μm范围内,并与其他控制方案相比,验证了所提方法的准确性和有效性。