非稀疏多核组合的支持向量回归方法

为了改善支持向量回归机的性能,本文提出了一种利用多核学习解决回归问题的算法(NS-MKR),算法对基本核函数的组合系数施加了Lp范数的约束(p>1),以得到组合系数的非稀疏解,并采用了两步优化方法,首先求解基于加权组合核的标准支持向量回归问题,用于学习拉格朗日乘子,然后采用简单的计算,求得基本核函数的组合系数,这两个步骤交替进行,直到满足事先定义的收敛准则。在人工数据集和真实数据集上的实验表明,相对于传统的单核和稀疏多核支持向量回归方法,本文提出的算法有更好的泛化性能。     

支持向量机的进化多核设计

为提高支持向量机分类精度,提出一种基于遗传程序设计的进化多核算法.算法中每个个体表示一个多核函数,并采用树形结构进行编码,增强了多核函数的非线性;初始种群由生长法产生,经过遗传操作后得到适合具体问题的进化多核函数.遗传程序设计的全局搜索性能使得算法设计不需要先验知识.与单核函数及其他多核函数的对比实验结果表明,进化多核...     

基于旋转SURF算子的图像配准新方法

针对SURF算法中快速Hessian矩阵行列式检测出的特征点的不连续现象,从而造成的旋转,模糊和光照变化适应性较差的不足,提出一种旋转SURF检测算子的图像配准新方法;该算法通过将SURF算法的积分图像盒子滤波模板逆时针旋转45度,引入一种可以检测角度旋转的滤波核提升检测算子对不同图像变换的匹配性能,保证新的检测算子与原算法较好的结合,同时利用改进的单纯形算法依据输入图像进行参数优化;仿真结果表明,该方法不仅保留了算法的速度优势,缩短了配准时间,而且在图像模糊变换,光照变换和JPEG压缩变换方面性能有明显的提升,此外对视角变换以及小尺度变换性能也有提高。     

An evolution strategy-based multiple kernels multi-criteria programming approach: The case of credit decision making

Credit risk analysis has long attracted a great deal of attention from both academic researchers and practitioners. However, because of the recent financial crisis, this field continues to draw ever increasingly attention. A multiple kernels multi-criteria programming approach based on evolution strategy (ES-MK-MCP) is proposed for credit decision making in this study. We introduce a linear combination of kernel functions to enhance the interpretability of credit classification models, and propose an alternative to optimize the parameters based on the evolution strategy. For illustration purpose, two UCI credit card data sets are used to verify the effectiveness and feasibility of the proposed model. As the experimental results reveal, the proposed ES-MK-MCP model is an efficient tool for credit risk analysis, especially for decision makers to identify the most relevant features.     

Sparse least squares support vector training in the reduced empirical feature space

In this paper we discuss sparse least squares support vector machines (sparse LS SVMs) trained in the empirical feature space, which is spanned by the mapped training data. First, we show that the kernel associated with the empirical feature space gives the same value with that of the kernel associated with the feature space if one of the arguments of the kernels is mapped into the empirical feature space by the mapping function associated with the feature space. Using this fact, we show that training and testing of kernel-based methods can be done in the empirical feature space and that training of LS SVMs in the empirical feature space results in solving a set of linear equations. We then derive the sparse LS SVMs restricting the linearly independent training data in the empirical feature space by the Cholesky factorization. Support vectors correspond to the selected training data and they do not change even if the value of the margin parameter is changed. Thus for linear kernels, the number of support vectors is the number of input variables at most. By computer experiments we show that we can reduce the number of support vectors without deteriorating the generalization ability.

基于三角形滤波卷积核的快速线积分卷积算法

基于纹理的线积分卷积(LIC)方法,能连续细致、动画地表示2D矢量大小和方向,但卷积计算极其耗时。针对这个问题实现了基于常数滤波卷积核的快速线积分卷积算法,提出并论证了基于三角形滤波卷积核的快速线积分卷积算法,对于实际矢量场,上述两种方法均将运算速度提高了一个数量级,而基于三角形滤波卷积核的快速线积分卷积算法的结果使图像更加平滑。     

A state space differential reproducing kernel method for the 3D analysis of FGM sandwich circular hollow cylinders with combinations of simply-supported and clamped edges

A state space differential reproducing kernel (DRK) method is developed for the three-dimensional (3D) analysis of functionally graded material (FGM) sandwich circular hollow cylinders with combinations of simply-supported and clamped edges and under sinusoidally (or uniformly) distributed loads. The strong formulation of this 3D elasticity problem is derived on the basis of the Reissner mixed variational theorem (RMVT), which consists of the Euler–Lagrange equations of this problem and its associated boundary conditions. The primary field variables are expanded as the single Fourier series in the circumferential coordinate, then interpolated in the axial coordinate using the early proposed DRK interpolation functions, and finally the state space equations of this problem are obtained, which represent a system of ordinary differential equations in the thickness coordinate. The present state space DRK solutions can then be obtained by means of the transfer matrix method. In the illustrative examples, three different edge conditions, the simple-simple (SS), simple-clamped (SC), and clamped–clamped (CC) edges, are considered, and the accuracy and convergence of this method are examined by comparing their solutions with the exact 3D ones available in the literature and the solutions using the ANSYS commercial software.     

Unsound kernel identification using linear colour charge‐coupled device

Wheat is one of the most consumed grains in the world. The identification of wheat based on surface characteristics is important for the market. This study is aimed at identifying unsound kernels (Triticum durum Desf), including 710 black germ kernels, 627 broken kernels and 1169 sound kernels from several seed distributors in China. The system is mainly composed of a liner charge‐coupled device for image capture and a software package for extracting various morphological, colour and texture features. The models built by partial least squares discriminate analysis, support vector machine discrimination analysis (SVMDA) and principal component analysis‐artificial neural networks for identifying the unsound kernels have been explored. After comparisons of these three methods, it has been found that SVMDA got the best accuracy: 95.1%, 96.0% and 98.3% (black germ kernels, broken kernels and sound kernels). Obviously, the experimental results have shown that SVMDA is the most feasible and effective choice for the identification.     

The journey of graph kernels through two decades

In the real world all events are connected. There is a hidden network of dependencies that governs behavior of natural processes. Without much argument it can be said that, of all the known data-structures, graphs are naturally suitable to model such information. But to learn to use graph data structure is a tedious job as most operations on graphs are computationally expensive, so exploring fast machine learning techniques for graph data has been an active area of research and a family of algorithms called kernel based approaches has been famous among researchers of the machine learning domain. With the help of support vector machines, kernel based methods work very well for learning with Gaussian processes. In this survey we will explore various kernels that operate on graph representations. Starting from the basics of kernel based learning we will travel through the history of graph kernels from its first appearance to discussion of current state of the art techniques in practice.