The Computational Complexity of the Chow Form

We present a bounded probability algorithm for the computation of the Chowforms of the equidimensional components of an algebraic variety. In particular, this gives an alternative procedure for the effective equidimensional decomposition of the variety, since each equidimensional component is characterized by its Chow form. The expected complexity of the algorithm is polynomial in the size and the geometric degree of the input equation system defining the variety. Hence it improves (or meets in some special cases) the complexity of all previous algorithms for computing Chow forms. In addition to this, we clarify the probability and uniformity aspects, which constitutes a further contribution of the paper. The algorithm is based on elimination theory techniques, in line with the geometric resolution algorithm due to M. Giusti, J. Heintz, L. M. Pardo, and their collaborators. In fact, ours can be considered as an extension of their algorithm for zero-dimensional systems to the case of positive-dimensional varieties. The key element for dealing with positive-dimensional varieties is a new Poisson-type product formula. This formula allows us to compute the Chow form of an equidimensional variety from a suitable zero-dimensional fiber. As an application, we obtain an algorithm to compute a subclass of sparse resultants, whose complexity is polynomial in the dimension and the volume of the input set of exponents. As another application, we derive an algorithm for the computation of the (unique) solution of a generic overdetermined polynomial equation system.     

Sparse-grid finite-volume multigrid for 3D-problems

We introduce a multigrid algorithm for the solution of a second order elliptic equation in three dimensions. For the approximation of the solution we use a partially ordered hierarchy of finite-volume discretisations. We show that there is a relation with semicoarsening and approximation by more-dimensional Haar wavelets. By taking a proper subset of all possible meshes in the hierarchy, a sparse grid finite-volume discretisation can be constructed.The multigrid algorithm consists of a simple damped point-Jacobi relaxation as the smoothing procedure, while the coarse grid correction is made by interpolation from several coarser grid levels.The combination of sparse grids and multigrid with semi-coarsening leads to a relatively small number of degrees of freedom,N, to obtain an accurate approximation, together with anO(N) method for the solution. The algorithm is symmetric with respect to the three coordinate directions and it is fit for combination with adaptive techniques.To analyse the convergence of the multigrid algorithm we develop the necessary Fourier analysis tools. All techniques, designed for 3D-problems, can also be applied for the 2D case, and — for simplicity — we apply the tools to study the convergence behaviour for the anisotropic Poisson equation for this 2D case.     

On solving sparse band systems with three algorithms of the ABS family

In this paper, we examine three algorithms in the ABS family and consider their storage requirements on sparse band systems. It is shown that, when using the implicit Cholesky algorithm on a band matrix with band width 2q+1, onlyq additional vectors are required. Indeed, for any matrix with upper band widthq, onlyq additional vectors are needed. More generally, ifa _kj 0,j>k, then thejth row ofH _i is effectively nonzero ifj>i>k. The arithmetic operations involved in solving a band matrix by this method are dominated by (1/2)n ² q. Special results are obtained forq-band tridiagonal matrices and cyclic band matrices.The implicit Cholesky algorithm may require pivoting if the matrixA does not possess positive-definite principal minors, so two further algorithms were considered that do not require this property. When using the implicit QR algorithm, a matrix with band widthq needs at most 2q additional vectors. Similar results forq-band tridiagonal matrices and cyclic band matrices are obtained.For the symmetric Huang algorithm, a matrix with band widthq requiresq–1 additional vectors. The storage required forq-band tridiagonal matrices and cyclic band matrices are again analyzed.This work was undertaken during the visit of Dr. J. Abaffy to Hatfield Polytechnic, sponsored by SERC Grant No. GR/E-07760.     

Sparse non-negative matrix factorizations for ultrasound factor analysis

Factor analysis is a powerful tool used for the analysis of dynamic studies. One of the major drawbacks of Factor Analysis of Dynamic Structures (FADS) is that the solution is not mathematically unique when only non-negativity constraints are used to determine factors and factor coefficients. In this paper, we introduce a novel method to correct FADS solutions by constructing and minimizing a new objective function. The method is improved from non-negative matrix factorizations (NMFs) algorithm by adding a sparse constraint that penalizes multiple components in the images of the factor coefficients. The technique is tested on computer simulations, and a patient ultrasound liver study. The results show that the method works well in comparison to the truth in computer simulations and to region of interest (ROI) measurements in the experimental studies.     

Localization of diffusion sources in complex networks with sparse observations

Locating sources in a large network is of paramount importance to reduce the spreading of disruptive behavior. Based on the backward diffusion-based method and integer programming, we propose an efficient approach to locate sources in complex networks with limited observers. The results on model networks and empirical networks demonstrate that, for a certain fraction of observers, the accuracy of our method for source localization will improve as the increase of network size. Besides, compared with the previous method (the maximum–minimum method), the performance of our method is much better with a small fraction of observers, especially in heterogeneous networks. Furthermore, our method is more robust against noise environments and strategies of choosing observers.     

Going Off the Grid: Iterative Model Selection for Biclustered Matrix Completion

We consider the problem of performing matrix completion with side information on row-by-row and column-by-column similarities. We build upon recent proposals for matrix estimation with smoothness constraints with respect to row and column graphs. We present a novel iterative procedure for directly minimizing an information criterion to select an appropriate amount of row and column smoothing, namely, to perform model selection. We also discuss how to exploit the special structure of the problem to scale up the estimation and model selection procedure via the Hutchinson estimator, combined with a stochastic Quasi-Newton approach. Supplementary material for this article is available online.     

TWO-SCALE FINITE ELEMENT DISCRETIZATIONS FOR PARTIAL DIFFERENTIAL EQUATIONS

Some two-scale finite element discretizations are introduced for a class of linear partialdifferential equations. Both boundary value and eigenvalue problems are studied. Basedon the two-scale error resolution techniques, several two-scale finite element algorithmsare proposed and analyzed. It is shown that this type of two-scale algorithms not onlysignificantly reduces the number of degrees of freedom but also produces very accurateapproximations.     

A bump triangular dynamic factorization algorithm for the simplex method

A dynamic factorization algorithm is developed which algebraically partitions the basis inverse in such a manner so that the simplex method can be executed from a series of small inverses and the basis itself. This partition is maintained dynamically so that the additional memory required to represent the basis inverse reduces to this series of small inverses for in-core implementations.The algorithm is intended for use in solving general large-scale linear programming problems. This new method of basis representation should permit rather large problems to be solved completely in-core.Preliminary computational experience is presented and comparisons are made with Reid's sparsity-exploiting variant of the Bartels—Golub decomposition for linear programming bases. The computational experience indicated that a significant reduction in memory requirements can usually be obtained using the dynamic factorization approach with only a slight (up to about 20%) degradation of execution time.This research was supported in part by the Air Force Office of Scientific Research, Air Force System Command, USAF, under AFOSR Contract/Grant Number AFOSR-74-2715.