Typical Kernel Research Articles

Pointed Subspace Approach to Incomplete Data

Incomplete data are often represented as vectors with filled missing attributes joined with flag vectors indicating missing components. In this paper, we generalize this approach and represent incomplete data as pointed affine subspaces. This allows to perform various affine transformations of data, such as whitening or dimensionality reduction. Moreover, this representation preserves the information, which coordinates were missing. To use our representation in practical classification tasks, we embed such generalized missing data into a vector space and define the scalar product of embedding space. Our representation is easy to implement, and can be used together with typical kernel methods. Performed experiments show that the application of SVM classifier on the proposed subspace approach obtains highly accurate results.

Pseudolabel Guided Kernel Learning for Hyperspectral Image Classification

In this paper, we propose a new framework for hyperspectral image classification, namely pseudolabel guided kernel learning (PLKL). The proposed framework is capable of fully utilizing unlabeled samples, making it very effective to handle the task with extremely limited training samples. Specifically, with multiple initial kernels and labeled samples, we first employ support vector machine (SVM) classifiers to predict pseudolabels independently for each unlabeled sample, and consistency voting is applied to the resulting pseudolabels to select and add a few unlabeled samples to the training set. Then, we refine the kernels to improve their discriminability with the augmented training set and a typical kernel learning method. Such phases are repeated until stable. Furthermore, we enhance the PLKL in terms of both the computation and memory efficiencies by using a bagging-like strategy, improving its practicality for large scale datasets. In addition, the proposed framework is quite flexible and general. That is, other advanced kernel-based methods can be incorporated to continuously improve the performance. Experimental results show that the proposed frameworks achieve much higher classification accuracy, compared with state-of-the-art methods. Especially, the classification accuracy improves more than 5% with very few training samples.

Synthesis of Linux Kernel Fuzzing Tools Based on Syscall

Any software especially the operating system requires testing and evaluation to validate the functional and security characteristics. As yet, fuzzing has become widely adopted into practice software testing. We focus on the operating system kernel fuzzing, select three typical kernel fuzzing tools to analyze. We make synthesis of the three kernel fuzzing tools from the aspects of sisal arguments model, test case construction and fuzzing scheduling and propose an abstract of partial specialization of model and explain it through these tools. Meanwhile, we inspect these tools on the usage of coverage-based fuzzing which is the state-of-the-art fuzzing optimization technology.

The landscape of GPGPU performance modeling tools

Sketch and taxonomies of current performance modeling landscape for GPGPU.A thorough description of 10 different approaches to GPU performance modeling.Empirical evaluation of models' performance using three kernels and four GPUs.Discussion of the strengths and weaknesses of the studied model classes. GPUs are gaining fast adoption as high-performance computing architectures, mainly because of their impressive peak performance. Yet most applications only achieve small fractions of this performance. While both programmers and architects have clear opinions about the causes of this performance gap, finding and quantifying the real problems remains a topic for performance modeling tools. In this paper, we sketch the landscape of modern GPUs' performance limiters and optimization opportunities, and dive into details on modeling attempts for GPU-based systems. We highlight the specific features of the relevant contributions in this field, along with the optimization and design spaces they explore. We further use typical kernel examples with various computation and memory access patterns to assess the efficacy and usability of a set of promising approaches. We conclude that the available GPU performance modeling solutions are very sensitive to applications and platform changes, and require significant efforts for tuning and calibration when new analyses are required.

Stationary Bootstrapping for the Nonparametric AR-ARCH Model

We consider a nonparametric AR(1) model with nonparametric ARCH(1) errors. In order to estimate the unknown function of the ARCH part, we apply the stationary bootstrap procedure, which is characterized by geometrically distributed random length of bootstrap blocks and has the advantage of capturing the dependence structure of the original data. The proposed method is composed of four steps: the first step estimates the AR part by a typical kernel smoothing to calculate AR residuals, the second step estimates the ARCH part via the Nadaraya-Watson kernel from the AR residuals to compute ARCH residuals, the third step applies the stationary bootstrap procedure to the ARCH residuals, and the fourth step defines the stationary bootstrapped Nadaraya-Watson estimator for the ARCH function with the stationary bootstrapped residuals. We prove the asymptotic validity of the stationary bootstrap estimator for the unknown ARCH function by showing the same limiting distribution as the Nadaraya-Watson estimator in the second step.

Towards consistence and convergence of conservative SPH approximations

Typical conservative smoothed particle hydrodynamics (SPH) approximations of the gradient of a scalar field introduce two errors: one (smoothing error) is due to smoothing of the gradient by an integration associated with a kernel function; the other (integration error) is due to approximating the integration by summation over all particles within the kernel support. When particles are not on a uniform grid, the integration error leads to violation of zero-order consistency, i.e. the inability to reproduce a constant field. In this paper we confirm that partition of unity is the condition under which the conservative SPH approximation achieves both consistence and convergence. We show that this condition can be achieved by relaxing a particle distribution under a constant pressure field and invariant particle volume. The resulting particle distribution is very similar to that is typical for liquid molecules. We further show that with two different typical kernel functions the SPH approximation, upon satisfying the partition of unity property, is able to achieve very high-order of the integration error, which previously could be shown only with particles on a uniform grid. The background pressure used in a weakly compressible SPH simulation implies a self-relaxation mechanism, which explains that convergence with respect to increasing particle numbers could be obtained in SPH simulations, although not predicted by previous numerical analysis. Furthermore, by relating the integration error to the background pressure, we explain why the previously proposed transport-velocity formulation of SPH (S. Adami et al. (2013) [1]) is able to achieve unprecedented accuracy and stability.

Investigation on Amari’s dynamical neural field with global constant inhibition

In this paper, the properties of Amari’s dynamical neural field with global constant inhibition induced by its kernel are investigated. Amari’s dynamical neural field illustrates many neurophysiological phenomena successfully and has been applied to unsupervised learning like data clustering in recent years. In its applications, the stationary solution to Amari’s dynamical neural field plays an important role that the underlying patterns being perceived are usually presented as the excited region in it. However, the type of stationary solution to dynamical neural field with typical kernel is often sensitive to parameters of its kernel that limits its range of application. Different from dynamical neural field with typical kernel that have been discussed a lot, there are few theoretical results on dynamical neural field with global constant inhibitory kernel that has already shown better performance in practice. In this paper, some important results on existence and stability of stationary solution to dynamical neural field with global constant inhibitory kernel are obtained. All of these results show that such kind of dynamical neural field has better potential for missions like data clustering than those with typical kernels, which provide a theoretical basis of its further extensive application.

Prediction of stress concentration factor of corrosion pits on buried pipes by least squares support vector machine

Concentrated working stress of a corrosion pit is known to be an important factor inducing the deterioration and consequently the breakage of a buried pipe. In this paper, we investigate the stress concentration factors (SCFs) of isolated elliptical corrosion pits using 3-D finite element analyses. The elliptical pits are fully characterized by their 3-D geometries, i.e., major and minor diameters and pit depth via a series of parametric studies of finite element pipe models. This is to quantify the effect of geometric variations of the pits on SCF. Realizing the fact that the 3-D finite element analysis is computationally intensive, efficient statistically predictive models have been developed based on least squares support vector machine (LS-SVM) realized in a ubiquitous spreadsheet platform. This approach shows very close predictions of SCF to the numerical findings for elliptical corrosion patterns on buried pipes. Particularly, two typical kernel functions have been adopted to cross validate the excellent performance of the LS-SVM method in SCF prediction of corroded pipes.

Interpretation of Helioseismic Travel Times

Time-distance helioseismology uses cross-covariances of wave motions on the solar surface to determine the travel times of wave packets moving from one surface location to another. We review the methodology to interpret travel-time measurements in terms of small, localized perturbations to a horizontally homogeneous reference solar model. Using the first Born approximation, we derive and compute 3D travel-time sensitivity (Fr\'echet) kernels for perturbations in sound-speed, density, pressure, and vector flows. While kernels for sound speed and flows had been computed previously, here we extend the calculation to kernels for density and pressure, hence providing a complete description of the effects of solar dynamics and structure on travel times. We treat three thermodynamic quantities as independent and do not assume hydrostatic equilibrium. We present a convenient approach to computing damped Green's functions using a normal-mode summation. The Green's function must be computed on a wavenumber grid that has sufficient resolution to resolve the longest lived modes. The typical kernel calculations used in this paper are computer intensive and require on the order of 600 CPU hours per kernel. Kernels are validated by computing the travel-time perturbation that results from horizontally-invariant perturbations using two independent approaches. At fixed sound-speed, the density and pressure kernels are approximately related through a negative multiplicative factor, therefore implying that perturbations in density and pressure are difficult to disentangle. Mean travel-times are not only sensitive to sound-speed, density and pressure perturbations, but also to flows, especially vertical flows. Accurate sensitivity kernels are needed to interpret complex flow patterns such as convection.

On Integral Invariants for Effective 3-D Motion Trajectory Matching and Recognition.

Motion trajectories tracked from the motions of human, robots, and moving objects can provide an important clue for motion analysis, classification, and recognition. This paper defines some new integral invariants for a 3-D motion trajectory. Based on two typical kernel functions, we design two integral invariants, the distance and area integral invariants. The area integral invariants are estimated based on the blurred segment of noisy discrete curve to avoid the computation of high-order derivatives. Such integral invariants for a motion trajectory enjoy some desirable properties, such as computational locality, uniqueness of representation, and noise insensitivity. Moreover, our formulation allows the analysis of motion trajectories at a range of scales by varying the scale of kernel function. The features of motion trajectories can thus be perceived at multiscale levels in a coarse-to-fine manner. Finally, we define a distance function to measure the trajectory similarity to find similar trajectories. Through the experiments, we examine the robustness and effectiveness of the proposed integral invariants and find that they can capture the motion cues in trajectory matching and sign recognition satisfactorily.

Image Classification Based on Histogram Intersection Kernel

Histogram Intersection Kernel Support Vector Machines (SVM) was used for the image classification problem. Specifically, each image was split into blocks, and each block was represented by the Scale Invariant Feature Transform (SIFT) descriptors; secondly, k-means cluster method was applied to separate the SIFT descriptors into groups, each group represented a visual keywords; thirdly, count the number of the SIFT descriptors in each image, and histogram of each image should be constructed; finally, Histogram Intersection Kernel should be built based on these histograms. In our experimental study, we use Corel-low images to test our method. Compared with typical RBF kernel SVM, the Histogram Intersection kernel SVM performs better than RBF kernel SVM.

Space Reconstruction of Industrial Heritage under Urban Transition - A Case Study of Hangzhou Section’s Renewal of the Grand Canal

On the basis the development transition of Hangzhou, this paper illustrates the industrial building resources in Hangzhou section based on the summary of practice pattern of domestic and foreign industrial heritage. It summarizes the typical kernel space feature of industrial heritage in Hangzhou section of the Grand Canal. Finally, the space reconstruction strategy of industrial heritage along the canal is proposed from the aspects of heritage patch, corridor, matrix trying to break the relatively isolated situation of the industrial heritage patches and urban matrix space.

KOMPSAT-2 영상의 토지피복분류에 적합한 SVM 커널 함수 비교 연구

Recently, the high-resolution satellite images is used the land cover and status data for the natural resources or environment management very helpful. The SVM algorithm of image processing has been used in various field. However, classification accuracy by SVM algorithm can be changed by various kernel functions and parameters. In this paper, the typical kernel function of the SVM algorithm was applied to the KOMPSAT-2 image and than the result of land cover performed the accuracy analysis using the checkpoint. Also, we carried out the analysis for selected the SVM kernel function from the land cover of the target region. As a result, the polynomial kernel function is demonstrated about the highest overall accuracy of classification. And that we know that the polynomial kernel and RBF kernel function is the best kernel function about each classification category accuracy.

Kernel linear regression for face recognition

Linear regression uses the least square algorithm to solve the solution of linear regression equation. Linear regression classification (LRC) shows good classification performance on face image data. However, when the axes of linear regression of class-specific samples have intersections, LRC could not well classify the samples that distribute around intersections. Moreover, the LRC could not perform well at the situation of severe lighting variations. This paper proposes a new classification method, kernel linear regression classification (KLRC), based on LRC and the kernel trick. KLRC is a nonlinear extension of LRC and can offset the drawback of LRC. KLRC implicitly maps the data into a high-dimensional kernel space by using the nonlinear mapping determined by a kernel function. Through this mapping, KLRC is able to make the data more linearly separable and can perform well for face recognition with varying lighting. For comparison, we conduct on three standard databases under some evaluation protocols. The proposed methodology not only outperforms LRC but also takes the better performance than typical kernel methods such as kernel linear discriminant analysis and kernel principal component analysis.

Power-Aware Parallel Loop Scheduling Method for Heterogeneous System

PDF HTML阅读 XML下载 导出引用 引用提醒 异构系统功耗感知的并行循环调度方法 DOI: 10.3724/SP.J.1001.2011.03883 作者: 作者单位: 作者简介: 通讯作者: 中图分类号: 基金项目: 国家自然科学基金(60921062, 60903059, 60903044); 国家科技重大专项(2009ZX01036-001-003-001) Power-Aware Parallel Loop Scheduling Method for Heterogeneous System Author: Affiliation: Fund Project: 摘要 | 图/表 | 访问统计 | 参考文献 | 相似文献 | 引证文献 | 资源附件 | 文章评论 摘要:以类OpenMP的并行程序为研究对象,在满足性能约束的条件下,结合异构系统并行循环调度和处理器动态电压调节技术优化系统功耗.首先建立了异构系统功耗感知的并行循环调度问题基本模型;然后,通过分析方法给出异构系统并行循环调度的能耗下界,该下界可用于评估功耗优化方法的实际效率;进而将异构系统并行循环调度问题归纳为整数规划问题,在此基础上,提出了处理器内循环再调度方法进一步降低功耗.最后,以CPU-GPU 异构系统为平台评测了10 个典型kernel 程序.实验结果表明,该方法可以有效降低系统功耗,提高系统效能. Abstract:Based on the OpenMP-like parallel program, a loop scheduling and dynamic voltage scaling technology is coordinated to optimize system power consumption under the given performance constraint. First, the basic model for power-aware loop scheduling on the heterogeneous system is presented. After that, through theoretical analysis, it has been concluded that the lower bound of energy consumption for parallel loop scheduling on heterogeneous systems, can be used as a baseline to evaluate the efficiency of optimization technology. Furthermore, this paper induces the scheduling problem as a typical integer programming problem and proposes inner-processor loop re-scheduling method to further reduce power consumption. Finally, 10 typical kernel programs on a CPU-GPU heterogeneous system are created. The experimental results demonstrate that the proposed method can effectively reduce the total energy consumption of the whole system and improve the system energy efficiency. 参考文献 相似文献 引证文献

Discriminant Graph Structures for Facial Expression Recognition

In this paper, a series of advances in elastic graph matching for facial expression recognition are proposed. More specifically, a new technique for the selection of the most discriminant facial landmarks for every facial expression (discriminant expression-specific graphs) is applied. Furthermore, a novel kernel-based technique for discriminant feature extraction from graphs is presented. This feature extraction technique remedies some of the limitations of the typical kernel Fisher discriminant analysis (KFDA) which provides a subspace of very limited dimensionality (i.e., one or two dimensions) in two-class problems. The proposed methods have been applied to the Cohn-Kanade database in which very good performance has been achieved in a fully automatic manner.

An Improved SVM Based on Similarity Metric

A novel support vector machine method for classification is presented in this paper. A modified kernel function based on the similarity metric and Rieman- nian metric is applied to the support vector machine. In general, it is believed that the similarity of homogeneous samples is higher than that of inhomogeneous samples. Therefore, in Riemannian geometry, Riemannian metric can be used to reflect local property of a curve. In order to enlarge the similarity metric of the homogeneous sam- ples or reduce that of the inhomogeneous samples in the feature space, Riemannian metric is used in the kernel function of the SVM. Simulated experiments are performed using the databases including an artificial and the UCI real data. Simulation results show the effectiveness of the proposed algorithm through the comparison with four typical kernel functions without similarity metric.

Role of Fungi Vectored by Rice Stink Bug (Heteroptera: Pentatomidae) in Discoloration of Rice Kernels

Fungi, yeast, and bacteria were cultured from field-collected, discolored (pecky) rice, Oryza sativa L., kernels and from stylets and saliva of the rice stink bug, Oebalus pugnax (F.). A sterilized chrome wire was used to wound and introduce selected inocula into developing kernels on rice panicles. Typical kernel discolorations could not be induced by inoculation with bacterial or yeast isolates. When the caryopsis was wounded during inoculation, several fungi caused discoloration comparable to that of pecky kernels collected from rice fields. Fusarium oxysporum Schlect, caused the greatest discoloration followed by Cochliobolus miyabeanus (Ito & Kuribayashi) Drechsler ex Dastur, Curvularia lunata (Wakker) Boedijn, Alternaria alternate (Fr.) Keissler, and A. Padwickii (Ganguly) M. B. Ellis. Symptoms typifying classical fungal discoloration peaked 5 to 10 dafter anthesis, and classical insect damage symptoms peaked 15 dafter anthesis when rice stink bugs were caged on plants 0, 5, 10, or 15 dafter anthesis. In field tests, density of rice stink bug infestations was associated positively with incidence of discolorations historically attributed to insects and fungi. These data indicate that discolorations in pecky rice result from fungi being introduced at the time of feeding of O. pugnax by means of a loose vector relationship.

On the Factorization of Certain Kernels Arising in Functional Equations of the Wiener-Hopf Type

The Wiener-Hopf technique is one of the most powerful methods for analyzing the scattering and diffraction problems of electromagnetic or acoustic waves. The essential step for the solution contains the so-called factorization . This procedure needs to split certain analytic functions (kernels) into the multiplication form of two functions which have semi-infinite regions of regularity. In this paper, factorization of typical kernel functions which arise in the Wiener-Hopf equations concerned with dielectric objects, is discussed. By applying a general factorization theorem, exact integral representations of split functions are obtained and then some approximate formulas are derived via rigorous asymptotic procedure. As a result, it is pointed out that final approximate formulas are useful for high frequency range.