Submitted by Jaqueline Silva (jtas29@gmail.com) on 2016-09-12T20:46:48Z No. of bitstreams: 2 Dissertação - Abssan Matuzinhos de Moura - 2016.pdf: 1317960 bytes, checksum: d406a9bf2b4d0bbca0ad6e3b52da498d (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) / Approved for entry into archive by Jaqueline Silva (jtas29@gmail.com) on 2016-09-12T20:47:11Z (GMT) No. of bitstreams: 2 Dissertação - Abssan Matuzinhos de Moura - 2016.pdf: 1317960 bytes, checksum: d406a9bf2b4d0bbca0ad6e3b52da498d (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) / Made available in DSpace on 2016-09-12T20:47:11Z (GMT). No. of bitstreams: 2 Dissertação - Abssan Matuzinhos de Moura - 2016.pdf: 1317960 bytes, checksum: d406a9bf2b4d0bbca0ad6e3b52da498d (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) Previous issue date: 2016-04-29 / The gradient method is a classical optimization methods to minimize a function. This method deserves special mention for its simplicity and easy understanding. This work is based on the study of the gradient method with step size given by the variant Barzilai- Borwein. Our goal is to present the convergence of the method with this variant. First we will study the two-dimensional case, for strictly convex quadratic functions. In this case, besides obtaining the convergence of the method, we see that such convergence occurs with R-superlinear rate. In the final part of the work, we will study the method with the variant Barzilai-Borwein not necessarily quadratic functions, concluding that the method converges. / O Método Gradiente é um dos métodos clássicos de otimização para minimizar uma função. Esse método merece um destaque especial pela sua simplicidade e fácil compreensão. Este trabalho se baseia no estudo do Método Gradiente com tamanho do passo dado pela variante de Barzilai-Borwein. Nosso objetivo é apresentar a convergência do método com esta variante. Primeiro faremos o estudo no caso bidimensional, para funções quadráticas estritamente convexas. Neste caso, além de obtermos a convergência do método, veremos que tal convergência ocorre com taxa R-superlinear. Na parte final do trabalho, faremos o estudo do método com a variante de Barzilai-Borwein para funções não necessariamente quadráticas, concluindo que o método converge.

La première partie est consacrée à la résolution de systèmes linéaires. Le chapitre 1 expose des résultats théoriques et numériques sur les méthodes proposées par Altman et précise le lien avec les méthodes de Krylov. Le chapitre 2 utilise des techniques d'extrapolation introduites par Brezinski pour obtenir une estimation du vecteur erreur. Plusieurs méthodes de projection sont retrouvées et de nouvelles procédures d'accélération données. Dans la deuxième partie, une nouvelle stratégie inspirée de la méthode de Cauchy-Barzilai-Borwein permet de définir de nouveaux schémas résolvant des problèmes de point fixe. Des résultats numériques sur un problème de bifurcation et un théorème de convergence sont donnés. Les chapitres 4, 5 et 6 sont consacrés à l'accélération de l'algorithme EM utilisé pour calculer des estimateurs du maximum de vraisemblance. Une classe de schémas itératifs basés sur la stratégie précédente est présentée, un théorème de convergence et une application à un problème de tomographie sont donnés. La dernière partie, fruit d'un projet du cemracs 2003, traite d'un problème issu de la physique des plasmas : l'amélioration des Codes Particles in Cell à l'aide d'une reconstruction de la densité basée sur une méthode d'ondelettes et sa validation numérique.

.

This thesis is broadly concerned with online-optimizing anti-windup control. These are control structures that implement some online-optimization routines to compensate for the windup effects in constrained control systems. The first part of this thesis examines a general framework for analyzing robust preservation in anti-windup control systems. This framework - the robust Kalman conjecture - is defined for the robust Lur’e problem. This part of the thesis verifies this conjecture for first-order plants perturbed by various norm-bounded unstructured uncertainties. Integral quadratic constraint theory is exploited to classify the appropriate stability multipliers required for verification in these cases. The remaining part of the thesis focusses on accelerated gradient methods. In particular, tight complexity-certificates can be obtained for the Nesterov gradient method, which makes it attractive for implementation of online-optimizing anti-windup control. This part of the thesis presents a proposed algorithm that extends the classical Nesterov gradient method by using available secant information. Numerical results demonstrating the efficiency of the proposed algorithm are analysed with the aid of performance profiles. As the objective function becomes more ill-conditioned, the proposed algorithm becomes significantly more efficient than the classical Nesterov gradient method. The improved performance bodes well for online-optimization anti-windup control since ill-conditioning is common place in constrained control systems. In addition, this thesis explores another subcategory of accelerated gradient methods known as Barzilai-Borwein gradient methods. Here, two algorithms that modify the Barzilai-Borwein gradient method are proposed. Global convergence of the proposed algorithms for all convex functions is established by using discrete Lyapunov theorems.