具有(k,r)-正交的(g,f)-因子分解的子图

研究了图的正交因子分解，通过构造函数p(x)和q(x)，证明了(mg k，mf-k)-图具有子图，该图有(g，f)-因子分解与kr-星(k，r)-正交，从而推广了原晋江教授的关于(mg m-1，mf-m 1)-图，存在(g，f)-因子分解与星(m，r)-正交的结论．     

Parallel and Systolic Solution of Normalized Explicit Approximate Inverse Preconditioning

A new class of normalized approximate inverse matrix techniques, based on the concept of sparse normalized approximate factorization procedures are introduced for solving sparse linear systems derived from the finite difference discretization of partial differential equations. Normalized explicit preconditioned conjugate gradient type methods in conjunction with normalized approximate inverse matrix techniques are presented for the efficient solution of sparse linear systems. Theoretical results on the rate of convergence of the normalized explicit preconditioned conjugate gradient scheme and estimates of the required computational work are presented. Application of the new proposed methods on two dimensional initial/boundary value problems is discussed and numerical results are given. The parallel and systolic implementation of the dominant computational part is also investigated.     

A numerical reconstruction method in inverse elastic scattering

In this paper a new numerical method for the shape reconstruction of obstacles in elastic scattering is proposed. Initially, the direct scattering problem for a rigid body and the mathematical setting for the corresponding inverse one are presented. Inverse uniqueness issues for the general case of mixed boundary conditions on the boundary of our obstacle, which are valid for a rigid body as well are established. The inversion algorithm based on the factorization method is presented into a suitable form and a new numerical scheme for the reconstruction of the shape of the scatterer, using far-field measurements, is given. In particular, an efficient Tikhonov parameter choice technique, called Improved Maximum Product Criterion (IMPC) and its linchpin within the framework of the factorization method is exploited. Our regularization parameter is computed via a fast iterative algorithm which requires no a priori knowledge of the noise level in the far-field data. Finally, the effectiveness of IMPC is illustrated with various numerical examples involving a kite, an acorn, and a peanut-shaped object.     

Weighted archetypal analysis of the multi-element graph for query-focused multi-document summarization

Most existing research on applying the matrix factorization approaches to query-focused multi-document summarization (Q-MDS) explores either soft/hard clustering or low rank approximation methods. We employ a different kind of matrix factorization method, namely weighted archetypal analysis (wAA) to Q-MDS. In query-focused summarization, given a graph representation of a set of sentences weighted by similarity to the given query, positively and/or negatively salient sentences are values on the weighted data set boundary. We choose to use wAA to compute these extreme values, archetypes, and hence to estimate the importance of sentences in target documents set. We investigate the impact of using the multi-element graph model for query focused summarization via wAA. We conducted experiments on the data of document understanding conference (DUC) 2005 and 2006. Experimental results evidence the improvement of the proposed approach over other closely related methods and many of state-of-the-art systems.     

Simplification of I-Vector Extraction for Speaker Identification

The identity vector (i-vector) approach has been the state-of-the-art for text-independent speaker recognition,both identification and verification in recent years.An i-vector is a low-dimensional vector in the socalled total variability space represented with a thin and tall rectangular matrix.This paper introduces a novel algorithm to improve the computational and memory requirements for the application.In our method,the series of symmetric matrices can be represented by diagonal expression,sharing the same dictionary,which to some extent is analogous to eigen decomposition,and we name this algorithm Eigen decomposition like factorization (EDLF).Similar algorithms are listed for comparison,in the same condition,our method shows no disadvantages in identification accuracy.     

Monitoring multivariate process variability when sub-group size is small

The current practice of multivariate process variability monitoring, when sub-group size is small, has nothing to do with probability of false alarm (PFA). Consequently, the reliability of the existing control charts remains undetermined. In this article, we propose a control chart which is reliable, very sensitive to the change in variance for small or moderate correlation, and provides a root causes analysis of an out-of-control signal. To illustrate these advantages, an industrial example is presented and the results are compared with those issued from the existing methods.     

基于LBSN动态异构网络的时间感知兴趣点推荐

随着基于位置社交网络(Location-Based Social Network,LBSN)的快速发展,兴趣点(Point-Of-Interest,POI)推荐可以帮助人们发现有趣的并吸引人的位置。针对签到数据的稀疏性和用户兴趣的动态性等挑战性问题,提出了基于LBSN动态异构网络的时间感知兴趣点推荐算法。在LBSN异构网络模式中增加会话节点类型。通过动态元路径,在用户和兴趣点语义关系之间有效地融入时间信息、位置信息和社交信息等。设置了用户-兴趣点之间的动态元路径集,并提出了动态路径实例的偏好度计算方法。采用矩阵分解模型对不同动态偏好矩阵进行矩阵分解。根据不同动态元路径的用户特征矩阵和兴趣点特征矩阵,获取用户在目标时间访问兴趣点的推荐列表。实验结果表明,与其他兴趣点推荐方法相比,所提方法在兴趣点推荐精确度上取得了较好的推荐结果,具有良好的应用前景。     

基于非负矩阵分解的单通道故障特征分离方法

针对单通道振动信号的多特征分离问题,提出了一种基于正交非负矩阵分解的故障特征提取方法。首先,采用短时傅里叶变换,利用时频分布来描述信号中的局部故障特征,通过核心一致性指标评估子空间维数;然后,在幅值谱矩阵分解的基础上,通过正交性约束实现低维嵌入分量信息的分离,获取局部特征的准确描述;最后,采用相位恢复理论重构出特征波形,对仿真信号和滚动轴承故障数据进行了测试。结果表明,所提出的方法能利用单通道信号有效地分离出微弱的局部故障特征,为机械状态的早期故障诊断识别提供了一种有效手段。     

增强型高阶注意力因子分解机点击率预测模型

特征交互的建模对于推荐系统中预测用户的点击率至关重要。设计了增强型高阶注意力因子分解机模型EHAFM(enhanced high-order attentive factorization machine),其主要由Embedding层、显式特征交互层、输出层构成。在每个显式特征交互层中,通过聚合其他特征的表示来更新特征的表示,并针对无用特征交互对预测产生干扰的问题,提出了增强型元素级注意力机制,利用投影矩阵拓展特征表示空间,以增强注意力矩阵的学习能力。通过融合多个增强型元素级注意力头的信息,以解决模型泛化能力不足问题。通过堆叠显式特征交互层可以将特征表示更新到任意高阶,将高阶特征交互部分与一阶线性部分结合进行点击率预测。EHAFM模型在Criteo、Movielens-1M两个数据集上进行实验,结果表明相较基准模型在两个数据集上分别有0.21%和0.92%的AUC提升。