Depuis les années 90, la méthode de création télescopique de Zeilberger a joué un rôle important dans la preuve automatique d'identités mettant en jeu des fonctions spéciales. L'objectif de long terme que nous attaquons dans ce travail est l'obtension d'algorithmes et d'implantations rapides pour l'intégration et la sommation définies dans le cadre de cette création télescopique. Nos contributions incluent de nouveaux algorithmes pratiques et des critères théoriques pour tester la terminaison d'algorithmes existants. Sur le plan pratique, nous nous focalisons sur la construction de télescopeurs minimaux pour les fonctions rationnelles en deux variables, laquelle a de nombreuses applications en lien avec les fonctions algébriques et les diagonales de séries génératrices rationnelles. En considérant cette classe d'entrées contraintes, nous parvenons à mâtiner la méthode générale de création télescopique avec réduction bien connue d'Hermite, issue de l'intégration symbolique. En outre, nous avons obtenu pour cette sous-classe quelques améliorations des algorithmes classiques d'Almkvist et Zeilberger. Nos résultats expérimentaux ont montré que les algorithmes à base de réduction d'Hermite battent tous les autres algorithmes connus, à la fois en ce qui concerne la complexité au pire et en ce qui concerne les mesures de temps sur nos implantations. Sur le plan théorique, notre premier résultat est motivé par la conjecture de Wilf et Zeilberger au sujet des fonctions hyperexponentielles-hypergéométriques holonomes. Nous présentons un théorème de structure pour les fonctions hyperexponentielles-hypergéométriques de plusieurs variables, indiquant qu'une telle fonction peut s'écrire comme le produit de fonctions usuelles. Ce théorème étend à la fois le théorème d'Ore et Sato pour les termes hypergéométriques en plusieurs variables et le résultat récent par Feng, Singer et Wu. Notre second résultat est relié au problème de l'existence de télescopeurs. Dans le cas discret à deux variables, Abramov a obtenu un critère qui indique quand un terme hypergéométrique a un télescopeur. Des résultats similaires ont été obtenus pour le $q$-décalage par Chen, Hou et Mu. Ces résultats sont fondamentaux pour la terminaison des algorithmes s'inspirant de celui de Zeilberger. Dans les autres cas mixtes continus/discrets, nous avons obtenu deux critères pour l'existence de télescopeurs pour des fonctions hyperexponentielles-hypergéométriques en deux variables. Nos critères s'appuient sur une représentation standard des fonctions hyperexponentielles-hypergéométriques en deux variables, sur sur deux décompositions additives.

The main goal of the thesis is to construct explicit, s-stage, strong-stability-preserving (SSP) Hermite–Birkhoff (HB) time discretization methods of order p with nonnegative coefficients for the integration of hyperbolic conservation laws. The Shu–Osher form and the canonical Shu–Osher form by means of the vector formulation for SSP Runge–Kutta (RK) methods are extended to SSP HB methods. The SSP coefficients of k-step, s-stage methods of order p, HB(k,s,p), as combinations of k-step methods of order (p − 3) with s-stage explicit RK methods of order 4, and k-step methods of order (p-4) with s-stage explicit RK methods of order 5, respectively, for s = 4, 5,..., 10 and p = 4, 5,..., 12, are constructed and compared with other methods. The good efficiency gains of the new, optimal, SSP HB methods over other SSP methods, such as Huang’s hybrid methods and RK methods, are numerically shown by means of their effective SSP coefficients and largest effective CFL numbers. The formulae of these new, optimal methods are presented in their Shu–Osher form.

This thesis is comprised of three main parts: The Hermite-Hadamard inequality on discrete time scales, the fractional Hermite-Hadamard inequality, and Karush-Kuhn- Tucker conditions on higher dimensional discrete domains. In the first part of the thesis, Chapters 2 & 3, we define a convex function on a special time scale T where all the time points are not uniformly distributed on a time line. With the use of the substitution rules of integration we prove the Hermite-Hadamard inequality for convex functions defined on T. In the fourth chapter, we introduce fractional order Hermite-Hadamard inequality and characterize convexity in terms of this inequality. In the fifth chapter, we discuss convexity on n{dimensional discrete time scales T = T1 × T2 × ... × Tn where Ti ⊂ R , i = 1; 2,…,n are discrete time scales which are not necessarily periodic. We introduce the discrete analogues of the fundamental concepts of real convex optimization such as convexity of a function, subgradients, and the Karush-Kuhn-Tucker conditions. We close this thesis by two remarks for the future direction of the research in this area.

This thesis presents new numerical algorithms for approximating functions by trigonometric polynomials and trigonometric rational functions. We begin by reviewing trigonometric polynomial interpolation and the barycentric formula for trigonometric polynomial interpolation in Chapter 1. Another feature of this chapter is the use of the complex plane, contour integrals and phase portraits for visualising various properties and relationships between periodic functions and their Laurent and trigonometric series. We also derive a periodic analogue of the Hermite integral formula which enables us to analyze interpolation error using contour integrals. We have not been able to find such a formula in the literature. Chapter 2 discusses trigonometric rational interpolation and trigonometric linearized rational least-squares approximations. To our knowledge, this is the first attempt to numerically solve these problems. The contribution of this chapter is presented in the form of a robust algorithm for computing trigonometric rational interpolants of prescribed numerator and denominator degrees at an arbitrary grid of interpolation points. The algorithm can also be used to compute trigonometric linearized rational least-squares and trigonometric polynomial least-squares approximations. Chapter 3 deals with the problem of trigonometric minimax approximation of functions, first in a space of trigonometric polynomials and then in a set of trigonometric rational functions. The contribution of this chapter is presented in the form of an algorithm, which to our knowledge, is the first description of a Remez-like algorithm to numerically compute trigonometric minimax polynomial and rational approximations. Our algorithm also uses trigonometric barycentric interpolation and Chebyshev-eigenvalue based root finding. Chapter 4 discusses the Fourier-Padé (called trigonometric Padé) approximation of a function. We review two existing approaches to the problem, both of which are based on rational approximations of a Laurent series. We present a numerical algorithm with examples and compute various type (m, n) trigonometric Padé approximants.

Made available in DSpace on 2014-06-12T17:36:51Z (GMT). No. of bitstreams: 2 arquivo2624_1.pdf: 8516783 bytes, checksum: 6d0f8f4a723a2ea3cd31f060df4e74e5 (MD5) license.txt: 1748 bytes, checksum: 8a4605be74aa9ea9d79846c1fba20a33 (MD5) Previous issue date: 2011 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Uma série de normas sobre sistemas de qualidade, denominada ISO 9000, estimularam as indústrias mundiais a garantirem padrões rígidos para se tornarem competitivos no mercado internacional. Assim, obter um controle eficiente sobre as máquinas e processos é extremamente necessário para obtenção de um desempenho satisfatório no aspecto de sistemas de controle. Nesta dissertação, apresenta-se uma sistemática para solução de um conjunto de parâmetros do controlador proporcional-integral (PI) para estabilização de determinada planta, sobretudo a sistemas com retardo onde a complexidade do sistema aumenta. Assim, é apresentado o Teorema de Hermite Biehler, onde é possível encontrar a solução através de uma sistemática, de maneira contrária aos métodos convencionais, como o critério de Routh-Hurwitz que esbarra em desigualdades polinomiais de difícil solução. É mostrada, também, a modelagem matemática apresentando o Método do Relé, e que alguns sistemas típicos industriais podem ser caracterizados em três categorias. Por fim, realizá-se um experimento prático, de maneira a consolidar toda a abordagem da dissertação

In clustered survival data, unobservable cluster effects may exert powerful influences on the outcomes and thus induce correlation among subjects within the same cluster. The ordinary partial likelihood approach does not account for this dependence. Frailty models, as an extension to Cox regression, incorporate multiplicative random effects, called frailties, into the hazard model and have become a very popular way to account for the dependence within clusters. We particularly study the two-level nested lognormal frailty model and propose an estimation approach based on the complete data likelihood with frailty terms integrated out. We adopt B-splines to model the baseline hazards and adaptive Gauss-Hermite quadrature to approximate the integrals efficiently. Furthermore, in finding the maximum likelihood estimators, instead of the Newton-Raphson iterative algorithm, Gauss-Seidel and BFGS methods are used to improve the stability and efficiency of the estimation procedure. We also study competing risks models with missing cause of failure in the context of Cox proportional hazards models. For competing risks data, there exists more than one cause of failure and each observed failure is exclusively linked to one cause. Conceptually, the causes are interpreted as competing risks before the failure is observed. Competing risks models are constructed based on the proportional hazards model specified for each cause of failure respectively, which can be estimated using partial likelihood approach. However, the ordinary partial likelihood is not applicable when the cause of failure could be missing for some reason. We propose a weighted partial likelihood approach based on complete-case data, where weights are computed as the inverse of selection probability and the selection probability is estimated by a logistic regression model. The asymptotic properties of the regression coefficient estimators are investigated by applying counting process and martingale theory. We further develop a double robust approach based on the full data to improve the efficiency as well as the robustness.

La construction des courbes parallèles est fondamentale pour différentes applications en modélisation géométrique, telles que l'étude des trajectoires d'outils pour les machines à commande numérique ou pour la définition des zones de tolérance. En général, la courbe parallèle d'une courbe rationnelle n'est pas rationnelle, ce qui conduit à déterminer une approximation de cette courbe parallèle par une courbe spline. Récemment, J. C. Fiorot et T. Gensane et indépendamment H. Pottmann ont donné la forme générale de toutes les courbes rationnelles à parallèles rationnelles (courbes à hodographe pythagorien). Dans cette dernière famille figurent les quartiques de Tschirnhausen. Ces courbes ont même flexibilité que les coniques, leurs courbes parallèles sont rationnelles de degré quatre et sont exactement les développantes des cubiques de Tschirnhausen. En se basant sur cette caractérisation, nous présentons un algorithme d'approximation, avec un contact d'ordre deux, d'une courbe et de ses parallèles par des quartiques de Tschirnhausen préservant la variation de la courbure. Par ailleurs, le caractère judicieux de la représentation Bézier duale des courbes à hodographe pythagorien et de leurs parallèles, nous a permis de construire des ovales et des rosettes rationnelles à largeur constante qui jouent un rôle important en mécanique des cames. Enfin, suite aux travaux de H. Busemann et H. Guggenheimer sur la géométrie plane de Minkowski, nous généralisons la notion de courbes parallèles ainsi que les résultats de H. Pottmann (concernant la caractérisation Bézier duale et la caractérisation géométrique des courbes à hodographe pythagorien) au plan de Minkowski

The main goal of the thesis is to construct explicit, s-stage, strong-stability-preserving (SSP) Hermite–Birkhoff (HB) time discretization methods of order p with nonnegative coefficients for the integration of hyperbolic conservation laws. The Shu–Osher form and the canonical Shu–Osher form by means of the vector formulation for SSP Runge–Kutta (RK) methods are extended to SSP HB methods. The SSP coefficients of k-step, s-stage methods of order p, HB(k,s,p), as combinations of k-step methods of order (p − 3) with s-stage explicit RK methods of order 4, and k-step methods of order (p-4) with s-stage explicit RK methods of order 5, respectively, for s = 4, 5,..., 10 and p = 4, 5,..., 12, are constructed and compared with other methods. The good efficiency gains of the new, optimal, SSP HB methods over other SSP methods, such as Huang’s hybrid methods and RK methods, are numerically shown by means of their effective SSP coefficients and largest effective CFL numbers. The formulae of these new, optimal methods are presented in their Shu–Osher form.

High-frequency financial data are not only discretely sampled in time but the time separating successive observations is often random. We review the paper of Aït-Sahalia and Mykland (2003), that measure the effects of discreteness sampling and ignoring the randomness of the sampling for estimating the m.l.e of a continuous-time diffusion model. In that article, three different assumptions and restrict in one made on the sampling intervals, and the corresponding likelihood function, asymptotic normality, and covariance matrix are obtained. It is concluded that the effects due to discretely sampling are smaller than the effect of simply ignoring the sampling randomness. This study focuses on rechecking the results in the paper of A¡Lıt-Sahalia and Mykland (2003) including theory, simulation and application. We derive a different likelihood function expression from A¡Lıt-Sahalia and Mykland (2003)¡¦s result. However, the asymptotic covariance are consistent for both approaching in the O-U process. Furthermore, we conduct an empirical study on the high frequency transaction time data by using non-homogeneous Poisson Processes.

Area averaging is one of the techniques used to solve problems encountered in the transport of momentum, heat, and mass. The application of this technique simplifies the mathematical solution of the problem. However, it necessitates expressing the local value of the dependent variable and/or its derivative(s) on the system boundaries in terms of the averaged variable. In this study, these expressions are obtained by the two-point Hermite expansion and this approximate method is applied to some specific problems, such as, unsteady flow in a concentric annulus, unequal cooling of a long slab, unsteady conduction in a cylindrical rod with internal heat generation, diffusion of a solute into a slab from limited volume of a well-mixed solution, convective mass transport between two parallel plates with a wall reaction, convective mass transport in a cylindrical tube with a wall reaction, and unsteady conduction in a two -layer composite slab. Comparison of the analytical and approximate solutions is shown to be in good agreement for a wide range of dimensionless parameters characterizing each system.