This paper presents the gSOAP stub and skeleton compiler. The compiler provides a unique SOAP-to-C/C++ language binding for deploying C/C++ applications in SOAP Web Services, clients, peer-to-peer computing networks. enables integratation of (legacy) C/C++/Fortran codes, embedded systems, real-time software peers that share computational resources information with other SOAP-enabled applications, possibly across different platforms, environments, disparate organizations located behind...

Scientific and engineering research is becoming increasingly dependent upon the development implementation of efficient parallel algorithms on modern high-performance computers. Numerical linear algebra an indispensable tool in such this paper attempts to collect describe a selection some its more important algorithms. The purpose review current status provide overall perspective for solving dense, banded, or block-structured problems arising major areas direct solution systems, least...

In this paper, we address the problem of constructing a reduced order system minimal McMillan degree that satisfiesa set tangential interpolation conditions with respect to original under some mild conditions. The resulting transfer function appears be generically unique and present simple efficient technique construct interpolating system. This is generalization multipoint Padé which particularly suited handle multiinput multioutput systems.

This paper develops and analyzes a generalization of the Broyden class quasi-Newton methods to problem minimizing smooth objective function $f$ on Riemannian manifold. A condition vector transport retraction that guarantees convergence facilitates efficient computation is derived. Experimental evidence presented demonstrating value extension through superior performance for some problems compared existing BFGS methods, in particular those depend differentiated retraction.

Linear algebra algorithms based on the BLAS or ex tended do not achieve high performance mul tivector processors with a hierarchical memory system because of lack data locality. For such machines, block linear must be implemented in terms matrix-matrix primitives (BLAS3). Designing ef ficient for these architectures requires analysis behavior and resulting as func tion certain parameters. The identify limits improvement possible via blocking any contradictory trends that require trade-off...

No abstract available.

Abstract In this paper, a novel approach for quantifying the parametric uncertainty associated with stochastic problem output is presented. As Monte-Carlo and collocation methods, only point-wise evaluations of response surface are required allowing use legacy deterministic codes precluding need any dedicated code to solve uncertain interest. The new differs from these standard methods in that it based on ideas directly linked recently developed compressed sensing theory. technique allows...

Although linear representations are frequently used in image analysis, their performances seldom optimal specific applications. This paper proposes a stochastic gradient algorithm for finding of images use appearance-based object recognition. Using the nearest neighbor classifier, recognition performance function is specified and that maximize this sought. For solving optimization problem on Grassmann manifold, utilizing intrinsic flows introduced. Several experimental results presented to...

Several blind source separation algorithms obtain a separating matrix by computing the congruence transformation that "best" diagonalizes collection of covariance matrices. Recent methods avoid pre-whitening phase and directly attempt to compute non-orthogonal diagonalizing congruence. However, since magnitude sources is unknown, there fundamental indeterminacy on norm rows matrix. We show how this can be taken into account restricting oblique manifold. The geometry manifold developed...

Riemannian optimization is the task of finding an optimum a real-valued function defined on manifold. has been topic much interest over past few years due to many applications including computer vision, signal processing, and numerical linear algebra. The substantial background required successfully design apply algorithms significant impediment for potential users. Therefore, multiple packages, such as Manopt (in Matlab) Pymanopt Python), have developed. This article describes ROPTLIB, C++...

In this paper, a Riemannian BFGS method for minimizing smooth function on manifold is defined, based generalization of cautious update and weak line search condition. It proven that the converges (i) globally to stationary points without assuming objective be convex (ii) superlinearly nondegenerate minimizer. Using condition removes need information from differentiated retraction. The joint matrix diagonalization problem chosen demonstrate performance algorithms with various parameters,...

It has long been known that a single ordering of optimization phases will not produce the best code for every application. This phase problem can be more severe when generating embedded systems due to need meet conflicting constraints on time, size, and power consumption. Given many application developers are willing spend time tuning an application, we believe viable approach is allow developer steer process optimizing function. In this paper, describe support in VISTA, interactive...

Parallel supercomputers architectures with complex memory hierarchies or distributed systems have become very common. Unfortunately, the problems associated restructuring software to take advantage of these are not easily solved. This paper presents an overview some mathematical issues behind several and attempts give a brief look at potential solutions.

Circuit simulation is a very time-consuming and numerically intensive application, especially when the problem size large as in case of VLSI circuits. To improve performance circuit simulators without sacrificing accuracy, variety parallel processing algorithms have been investigated. Research field surveyed, ongoing research this area at University Illinois described. Both standard relaxation-based approaches are considered. In particular, forms parallelism available within direct method...