An integrated framework of learning and evidential reasoning for user profiling using short texts

Inferring user profiles based on texts created by users on social networks has a variety of applications in recommender systems such as job offering, item recommendation, and targeted advertisement. The problem becomes more challenging when working with short texts like tweets on Twitter, or posts on Facebook. This work aims at proposing an integrated framework based on Dempster–Shafer theory of evidence, word embedding, and $k$-means clustering for user profiling problem, which is capable of not only working well with short texts but also dealing with uncertainty inherently in user texts. The proposed framework is essentially composed of three phases: (1) Learning abstract concepts at multiple levels of abstraction from user corpora; (2) Evidential inference and combination for user modeling; and (3) User profile extraction. Particularly, in the first phase, a word embedding technique is used to convert preprocessed texts into vectors which capture semantics of words in user corpus, and then $k$-means clustering is utilized for learning abstract concepts at multiple levels of abstraction, each of which reflects appropriate semantics of user profiles. In the second phase, by considering each document in user corpus as an evidential source that carries some partial information for inferring user profiles, we first infer a mass function associated with each user document by maximum a posterior estimation, and then apply Dempster’s rule of combination for fusing all documents’ mass functions into an overall one for the user corpus. Finally, in the third phase, we apply the so-called pignistic probability principle to extract top-$n$ keywords from user’s overall mass function to define the user profile. Thanks to the ability of combining pieces of information from many documents, the proposed framework is flexible enough to be scaled when input data coming from not only multiple modes but different sources on web environments. Besides, the resulting profiles are interpretable, visualizable, and compatible in practical applications. The effectiveness of the proposed framework is validated by experimental studies conducted on datasets crawled from Twitter and Facebook.     

An accelerated cyclic-reduction-based solvent method for solving quadratic eigenvalue problem of gyroscopic systems

The quadratic eigenvalue problem (QEP) $({\lambda }^{2}M+\lambda G+K)x=0$, with $M^T=M$ being positive definite, $K^T=K$ being negative definite and $G^T=?G$, is associated with gyroscopic systems. In Guo (2004), a cyclic-reduction-based solvent (CRS) method was proposed to compute all eigenvalues of the above mentioned QEP. Firstly, the problem is converted to find a suitable solvent of the quadratic matrix equation (QME) $M{X}^2+GX+K=0$. Then using a Cayley transformation and a proper substitution, the QME is transformed into the nonlinear matrix equation (NME) $Z+A^T{Z}^{?1}A=Q$ with $A=M+K+G$ and $Q=2(M?K)$. The problem finally can be solved by applying the CR method to obtain the maximal symmetric positive definite solution of the NME as long as the QEP has no eigenvalues on the imaginary axis or for some cases where the QEP has eigenvalues on the imaginary axis. However, when all eigenvalues of the QEP are far away from or near the origin, the Cayley transformation seems not to be the best one and the convergence rate of the CRS method proposed in Guo (2004) might be further improved. In this paper, inspired by using a doubling algorithm to solve the QME, we use a Möbius transformation instead of the Cayley transformation to present an accelerated CRS (ACRS) method for solving the QEP of gyroscopic systems. In addition, we discuss the selection strategies of optimal parameter for the ACRS method. Numerical results demonstrate the efficiency of our method.     

GPU implementation of Borůvka’s algorithm to Euclidean minimum spanning tree based on Elias method

We present both sequential and data parallel approaches to build hierarchical minimum spanning forest (MSF) or tree (MST) in Euclidean space (EMSF/EMST) for applications whose input $N$ points are uniformly or boundedly distributed in Euclidean space. Each iteration of the sequential approach takes $O\left(N\right)$ time complexity through combining Borůvka’s algorithm with an improved component-based neighborhood search algorithm, namely sliced spiral search, which is a newly proposed improvement to Bentley’s spiral search for finding a component graph’s closest outgoing point on the plane. It works based on the uniqueness property in Euclidean space, and allows $O\left(1\right)$ time complexity for one search from a query point to find the component’s closest outgoing point at different iterations of Borůvka’s algorithm. The data parallel approach includes a newly proposed two-direction breadth-first search (BFS) implementation on graphics processing unit (GPU) platform, which is specialized for selecting a spanning tree’s shortest outgoing edge. This GPU two-direction parallel BFS enables a tree traversal operation to start from any one of its vertex acting as root. These GPU parallel implementations work by assigning $N$ threads with one thread associated to one input point, one thread occupies $O\left(1\right)$ local memory and the whole algorithm occupies $O\left(N\right)$ global memory. Experiments are conducted on both uniformly distributed data sets and TSPLIB database. We evaluate computation time of the proposed approaches on more than 80 benchmarks with size $N$ growing up to $10^6$ points on personal laptop.     

A remark concerning divergence accuracy order for H(div)-conforming finite element flux approximations

The construction of finite element approximations in $\mathbf{H}(\text{div},\Omega )$ usually requires the Piola transformation to map vector polynomials from a master element to vector fields in the elements of a partition of the region $\Omega$. It is known that degradation may occur in convergence order if non affine geometric mappings are used. On this point, we revisit a general procedure for the improvement of two-dimensional flux approximations discussed in a recent paper of this journal (Comput. Math. Appl. 74 (2017) 3283–3295). The starting point is an approximation scheme, which is known to provide $L^2$-errors with accuracy of order $k+1$ for sufficiently smooth flux functions, and of order $r+1$ for flux divergence. An example is $R{T}_k$ spaces on quadrilateral meshes, where $r=k$ or $k?1$ if linear or bilinear geometric isomorphisms are applied. Furthermore, the original space is required to be expressed by a factorization in terms of edge and internal shape flux functions. The goal is to define a hierarchy of enriched flux approximations to reach arbitrary higher orders of divergence accuracy $r+n+1$ as desired, for any $n\ge 1$. The enriched versions are defined by adding higher degree internal shape functions of the original family of spaces at level $k+n$, while keeping the original border fluxes at level $k$. The case $n=1$ has been discussed in the mentioned publication for two particular examples. General stronger enrichment $n>1$ shall be analyzed and applied to Darcy’s flow simulations, the global condensed systems to be solved having same dimension and structure of the original scheme.     

Matrix LSQR algorithm for structured solutions to quaternionic least squares problem

In this paper, we employ matrix LSQR algorithm to deal with quaternionic least squares problem in order to find the minimum norm solutions with kinds of special structures, and propose a strategy to accelerate convergence rate of the algorithm via right–left preconditioning of the coefficient matrices. We mainly focus on analyzing the minimum norm $\eta$-Hermitian solution and the minimum norm $\eta$-biHermitian solution to the quaternionic least squares problem, $\eta \in \{\mathbf{i},\mathbf{j},\mathbf{k}\}$. Other structured solutions also can be obtained using the proposed technique. A number of numerical experiments are performed to show the efficiency of the preconditioned matrix LSQR algorithm.     

A fast singular boundary method for 3D Helmholtz equation

This paper presents a fast singular boundary method (SBM) for three-dimensional (3D) Helmholtz equation. The SBM is a boundary-type meshless method which incorporates the advantages of the boundary element method (BEM) and the method of fundamental solutions (MFS). It is easy-to-program, and attractive to the problems with complex geometries. However, the SBM is usually limited to small-scale problems, because of the operation count of $O\left(N^3\right)$ with direct solvers or $O\left(N^2\right)$ with iterative solvers, as well as the memory requirement of $O\left(N^2\right)$. To overcome this drawback, this study makes the first attempt to employ the precorrected-FFT (PFFT) to accelerate the SBM matrix–vector multiplication at each iteration step of the GMRES for 3D Helmholtz equation. Consequently, the computational complexity can be reduced from $O\left(N^2\right)$ to $O\left(NlogN\right)$ or $O\left(N\right)$. Three numerical examples are successfully tested on a desktop computer. The results clearly demonstrate the accuracy and efficiency of the developed fast PFFT-SBM strategy.     

Stochastic 2D primitive equations: Central limit theorem and moderate deviation principle

In this paper, we establish a central limit theorem and a moderate deviation for two-dimensional stochastic primitive equations driven by multiplicative noise. This is the first result about the limit theorem and the moderate deviations for stochastic primitive equations. The proof of the results relies on the weak convergence method and some delicate and careful $a$ $priori$ estimates.     

Superconvergence analysis of nonconforming finite element method for time-fractional nonlinear parabolic equations on anisotropic meshes

In this paper, we prove a novel result of the consistency error estimate with order $O\left(h^2\right)$ for $E{Q}_1^{rot}$ element (see Lemma 2) on anisotropic meshes. Then, a linearized fully discrete Galerkin finite element method (FEM) is studied for the time-fractional nonlinear parabolic problems, and the superclose and superconvergent estimates of order $O(\tau +h^2)$ in broken $H^1$-norm on anisotropic meshes are derived by using the proved character of $E{Q}_1^{rot}$ element, which improve the results in the existing literature. Numerical results are provided to confirm the theoretical analysis.