Stacked Regressions

Stacking regressions is a method for forming linear combinations of different predictors to give improved prediction accuracy. The idea is to use cross-validation data and least squares under non negativity constraints to determine the coefficients in the combination. Its effectiveness is demonstrated in stacking regression trees of different sizes and in a simulation stacking linear subset and ridge regressions. Reasons why this method works are explored. The idea of stacking originated with Wolpert (1992).     

Is Combining Classifiers with Stacking Better than Selecting the Best One?

We empirically evaluate several state-of-the-art methods for constructing ensembles of heterogeneous classifiers with stacking and show that they perform (at best) comparably to selecting the best classifier from the ensemble by cross validation. Among state-of-the-art stacking methods, stacking with probability distributions and multi-response linear regression performs best. We propose two extensions of this method, one using an extended set of meta-level features and the other using multi-response model trees to learn at the meta-level. We show that the latter extension performs better than existing stacking approaches and better than selecting the best classifier by cross validation.     

结合特征图谱学习的人数统计方法

针对实际公共场景视频的人数统计中存在的背景干扰、光照变化、目标间遮挡等问题,提出一种结合特征图谱学习和一阶动态线性回归的人数统计方法。首先,建立图像的尺度不变特征变换（SIFT）特征与目标真实密度图之间的特征图谱映射模型,利用SIFT特征和前述映射模型得到包含目标和背景特征量的特征图谱;然后,根据通常监控视频中背景变化较小、特征图谱中的背景特征量相对稳定的特点,由特征图谱的积分与真实人数通过一阶动态线性回归建立人数回归模型;最后,通过该回归模型模型得出估计人数。在数据集MALL和PETS2009上进行实验,实验结果表明：与累积属性空间方法相比,所提方法平均绝对误差降低了2.2%;与基于角点检测的一阶动态线性回归方法相比,其平均绝对误差降低了6.5%,平均相对误差降低了2.3%。     

Prediction in evolutionary algorithms for dynamic environments

Evolutionary algorithms have been widely used to solve dynamic optimization problems. Memory-based evolutionary algorithms are often used when the dynamics of the environment follow some repeated behavior. Over the last few years, the use of prediction mechanisms combined with memory has been explored. These prediction techniques are used to avoid the decrease of the algorithm’s performance when a change occurs. This paper investigates the use of prediction methods in memory-based evolutionary algorithms for two distinct situations: to predict when the next change will happen and how the environment will change. For the first predictor two techniques are explored, one based on linear regression and another supported by nonlinear regression. For the second, a technique based on Markov chains is explored. Several experiments were carried out using different types of dynamics in two benchmark problems. Experimental results show that the incorporation of the proposed prediction techniques efficiently improves the performance of evolutionary algorithms in dynamic optimization problems.     

网格数据融合系统中的数据源管理

针对网格海量动态数据高效融合的要求,提出提出一个套新的数据源管理方案,主要包括一套一套面向对象的元数据描述规范和、基于此规范的数据查询机制以及、模式融合策略,以解决传统网格中动态异构数据源难以维护的问题。给出,一种基于聚类的有效类挖掘算法。实验证明,该能有效地解决传统网格中动态异构数据源难以维护的问题。另外提出一种基于聚类的有效类挖掘算法,实验证明算法能大大优化系统的性能。     

光频收发信机制及其在计算机中的应用

冯诺曼机器以提高处理器频率增强性能,多处理器环境当中的系统性能决定于使更多处理器协同工作。通信带宽是高性能计算系统的首要瓶颈。本文提出一种芯片间通信的神经脊,把数以百计的芯片连接成一组网络,为多处理单元并行运算提供一个骨架。用相干的激光作为通信媒介,以垂直腔面的激光元件组成收发信机。相干光充分利用谐振线路放大器的特性提升通信的扇出,令芯片间的信息传递效率大大提高。使通信网络的复杂性从按通信端点数n*(n+1)式上升,变为线性增加。     

A robust regression-based linear model adjustment method for SAR interferogram stacking techniques

Differential interferometric synthetic aperture radar (DInSAR) is recognized as an effective remote-sensing technique for a variety of ground deformation mapping applications. Centimetre-level measurement accuracy can be achieved with the DInSAR technique. However, two key limitations – temporal decorrelation and phase delay due to atmospheric inhomogeneities – might decrease the accuracy of deformation measurements. To overcome such problems, interferogram stacking techniques, which extend the DInSAR technique, have been developed in recent years. In most implementations of such techniques, a so-called ‘linear model adjustment’ step is required to obtain the relative linear deformation rate and digital elevation model error from the double-differenced phase observations along the stack. In this step, since a non-linear system has to be resolved, the traditional least squares method cannot be directly applied. In order to overcome this problem, several methods have been developed in recent years. In this article, a new method has been developed to deal with the problem of linear model adjustment. This method repeatedly uses robust regression to resolve the non-linear system and is much easier to implement compared with other methods. This method is applied to both simulated and real data, and the results demonstrate that it can be efficiently used for linear model adjustment.     

基于动态射频指纹的调频定位方法

针对调频(FM)广播信号动态变化对定位性能产生较大影响的问题,提出了动态射频指纹的FM定位方法。该方法采用多元线性回归和神经网络方法,根据离线阶段各个参考点和部分较准点的调频广播信号强度的关系,使用在线阶段较准点的信号强度实时估计在线阶段参考点的信号强度。通过这两种方法建立具有自适应能力的动态射频地图,并使用贝叶斯估计方法对目标进行定位。实验结果表明,相对静态射频指纹模型：采用多元线性回归的射频指纹动态映射模型定位误差平均减少9.1%,采用神经网络的射频指纹动态映射模型定位误差平均减少36.3%,有效抑制了射频信号动态变化时变性对定位性能的影响。     

A survey of memory architecture for 3D chip multi-processors

3D chip multi-processors (3D CMPs) combine the advantages of 3D integration and the parallelism of CMPs, which are emerging as active research topics in VLSI and multi-core computer architecture communities. One significant potentiality of 3D CMPs is to exploit the diversity of integration processes and high volume of vertical TSV bandwidth to mitigate the well-known “Memory Wall” problem. Meanwhile, the 3D integration techniques are under the severe thermal, manufacture yield and cost constraints. Research on 3D stacking memory hierarchy explores the high performance and power/thermal efficient memory architectures for 3D CMPs. The micro-architectures of memories can be designed in the 3D integrated circuit context and integrated into 3D CMPs. This paper surveys the design of memory architectures for 3D CMPs. We summarize current research into two categories: stacking cache-only architectures and stacking main memory architectures for 3D CMPs. The representative works are reviewed and the remaining opportunities and challenges are discussed to guide the future research in this emerging area.     

Nationwide Freight Generation Models: A Spatial Regression Approach

This paper investigates the application of linear regression models and modeling techniques in predicting freight generation at the national level within the U.S. Specifically, the paper seeks to improve the performance and fit of linear regression models of freight generation. We provide insight into different variable transformation techniques, evaluate the use of spatial regression variables, and apply a spatial regression modeling methodology to correct for spatial autocorrelation. We conclude that the spatial regression model is the preferred specification for freight generation at the national level. The proliferation of Geographic Information Systems (GIS) within planning agencies affords more widespread use of spatial regression and our results indicate this technique would provide improvement to models that have been traditionally limited by insufficient data.     

基于滑动平均与分段线性回归的时间序列相似性

针对时间序列相似性度量中欧氏距离对异常数据敏感以及DTW距离算法效率低的问题,提出基于滑动平均与分段线性回归的时间序列相似性方法。首先,使用初始可变滑动平均算法以及分段线性回归对原始时间序列进行数据变换,并将分段线性回归的参数(截距与距离)集作为时间序列的特征,以实现时间序列的特征提取和数据降维;然后,利用动态时间弯曲距离进行距离计算。该方法在时间序列相似性上与DTW算法的性能相近,但是在算法效率上几乎提高了96%。实验结果验证了该方法的有效性与准确性。     

Sparse identification for ball-screw drives considering position-dependent dynamics and nonlinear friction

Establishing accurate dynamic models in a form that is suitable for integration with model-based control methods, is of great significance for further improving the dynamic motion control precision of ball-screw drives. However, due to the nonlinear time-varying factors such as position-dependent dynamics and nonlinear friction disturbance, it is difficult to model the dynamic characteristics of ball-screw drives accurately, concisely and efficiently. To overcome this challenge, a sparse identification method for ball-screw drives is proposed. Ball-screw drives are modeled as discrete-time linear parameter-varying systems under nonlinear friction disturbance, and two types of dictionary function libraries are designed to represent the position-dependent dynamics and nonlinear friction respectively. After constructing the regression form of the system model, a stepwise sparse regression policy is proposed to solve all the coefficients of dictionary functions. The proposed method is verified in both simulation and real environments. The results both show that by the proposed method, an accurate and linearizable dynamic model of ball-screw drives can be identified only using the data from only one global random excitation experiment covering the working stroke.     

应用DRNN和ARIMA组合模型的时空集成预测方法

在评析目前时空预测研究现状的基础上,提出基于动态回归神经网络(DRNN)和自回归集成移动平均(ARIMA)组合模型的时空集成预测方法.该方法先用ARIMA模型对时空数据的时序进行预测,再用DRNN捕获时空数据间隐藏的空间关系,最后用线性回归将二者整合起来,得到集成预测结果.案例实验结果表明:该方法比不考虑空间影响的预测方法或单一的预测方法有更高的精度;该方法具有良好的动态处理和计算能力,对跨空间的动态过程的预测有效可行.     

半在线入库堆垛问题的动态求解算法

堆场垛位优化问题一直是仓储管理的难点和焦点之一,垛位优化可以保证物料装卸和出入库的高效率,同时对保证合同交货期也起着至关重要的作用。针对仓储和生产一体化下的入库堆垛问题,本文通过分析将其归结为一类半在线的A型装箱问题,并依据问题的特点,建立了最小化总倒垛次数的优化模型。根据货场天车在相邻入库过程中存在空闲作业量的特点,设计了一种前序货物允许移动的动态堆垛策略,结合堆垛约束后嵌入到经典装箱启发式算法中,最后通过仿真算例验证了该策略的有效性。     

非线性动态系统中迭代的同步数据协调与显著误差检测的支持向量回归方法

化工厂中过程数据的质量严重影响到来自例如性能监控, 在线优化和控制等活动所获得的效益和性能. 由于许多化工过程通常表现为非线性动态特性, 例如扩展卡尔曼滤波(EKF)和非线性动态数据协调(NDDR)等技术已经被发展出来改进数据的质量. 近期, 迭代非线性动态数据协调(RNDDR)技术已被提出, 该技术结合了EKF和NDDR技术的优点. 但是, RNDDR技术不能够处理具有显著误差的测量值. 本文中, 一种非线性动态系统中迭代的同步数据协调与显著误差检测的支持向量(SV)回归方法被提出. SV回归是一种经验风险和结构风险间的妥协, 并且对于数据协调来说, 其对随机误差和显著误差是鲁棒的．通过将结构风险取代RNDDR中的极大似然估计并使其最小化, 我们的方法不仅可以实现迭代非线性动态数据协调, 还可以同时实现显著误差检测. 本文中的非线性动态系统仿真结果显示出, 所提出的方法在迭代实时估计框架下, 对于非线性动态系统的同步数据协调和显著误差检测是鲁棒、稳定并且精确的. 该方法也可以提供更好的控制性能.     

Linear combination of component results in information retrieval

In information retrieval, data fusion (also known as meta-search) has been investigated by many researchers. Previous investigation and experimentation demonstrate that the linear combination method is an effective data fusion method for combining multiple information retrieval results. One advantage is its flexibility, since different weights can be assigned to different component systems so as to obtain better fusion results. The key issue is how to assign good weights to all the component retrieval systems involved. Surprisingly, research in this field is limited and it is still an open question. In this paper, we use the multiple linear regression technique with estimated relevance scores and judged scores to obtain suitable weights. Although the multiple linear regression technique is not new, the way of using it in this paper has never been attempted before for the data fusion problem in information retrieval. Our experiments with five groups of runs submitted to TREC show that the linear combination method with such a weighting strategy steadily outperforms the best component system and other data fusion methods including CombSum, CombMNZ, PosFuse, MAPFuse, SegFuse, and the linear combination method with performance level/performance square weighting schemes by large margins.     

Genetic algorithms-based fuzzy regression analysis

 This paper describes the concept of fuzzy regression analysis based on genetic algorithms. It is shown that the performance of fuzzy regression models may be improved and fuzzy modeling technique can be simplified by incorporating genetic algorithms into regression analysis procedure. The effectiveness of the proposed approach is illustrated through simulation of fuzzy linear regression model obtained by other authors and comparison of the results. The paper further demonstrates the applications of the approach to the manufacturing and business problems.     

Bias reduction in kernel binary regression

We consider Nadaraya-Watson type estimators for binary regression functions. We propose a method for improving the performance of such estimators by employing bias reduction techniques when estimating the constituent probability densities. Direct substitution of separately optimized density estimates into the regression function formula generates disappointing results in practice. However, adjusting the global smoothing parameter to optimize a performance criterion for the binary regression function itself is more promising. We focus on an implementation of this approach which uses a variable kernel technique to provide reduced bias density estimates, and where the global bandwidth is selected by an appropriately tailored leave-one-out (cross-validation) method. Theory and numerical experiments show that this form of bias reduction improves performance substantially when the underlying regression function is highly non-linear but is not beneficial when the underlying regression function is almost linear in form.     

CATC — A Computer Aided Tolerance Control system

Computer utilization in manufacturing systems is increasing at a rapid rate. Developments thus far have been disjointed. Much work needs to be done in the area of integrating computer aided design and computer aided manufacturing (CAD/CAM). Tolerance control is an important component in the integration of CAD and CAM. Selection of design tolerances affects the build-ability of a product. The selection of manufacturing processes and sequence of processes affects process tolerance stacking. Unexamined process tolerance stacking leads to scrap and rework. A computer aided tolerance control (CATC) system is presented in this paper. This system is based on the tolerance chart technique. The CATC system is interactive and uses computer graphics for information display. The system can be used for computer aided process planning (CAPP) and for CAD/CAM integration.     

Degradation of linear filter performance due to modeling error

Differential equations for determining the dynamic and steady-state effects of a particular class of disturbances on the error in the estimate of the state vector of a stochastic linear dynamic system are obtained. For the problem of evaluating near optimal filter performance, the technique permits the performance degradation due to the deletion of certain state vector components in the design of a Kalman filter to be obtained.