No description available.

Many problems in signal processing, machine learning and computer vision can be solved by learning low rank models from data. In computer vision, problems such as rigid structure from motion have been formulated as an optimization over subspaces with fixed rank. These hard-rank constraints have traditionally been imposed by a factorization that parameterizes subspaces as a product of two matrices of fixed rank. Whilst factorization approaches lead to efficient and kernelizable optimization algorithms, they have been shown to be NP-Hard in presence of missing data. Inspired by recent work in compressed sensing, hard-rank constraints have been replaced by soft-rank constraints, such as the nuclear norm regularizer. Vis-a-vis hard-rank approaches, soft-rank models are convex even in presence of missing data: but how is convex optimization solving a NP-Hard problem? This thesis addresses this question by analyzing the relationship between hard and soft rank constraints in the unsupervised factorization with missing data problem. Moreover, we extend soft rank models to weakly supervised and fully supervised learning problems in computer vision. There are four main contributions of our work: (1) The analysis of a new unified low-rank model for matrix factorization with missing data. Our model subsumes soft and hard-rank approaches and merges advantages from previous formulations, such as efficient algorithms and kernelization. It also provides justifications on the choice of algorithms and regions that guarantee convergence to global minima. (2) A deterministic \rank continuation" strategy for the NP-hard unsupervised factorization with missing data problem, that is highly competitive with the state-of-the-art and often achieves globally optimal solutions. In preliminary work, we show that this optimization strategy is applicable to other NP-hard problems which are typically relaxed to convex semidentite programs (e.g., MAX-CUT, quadratic assignment problem). (3) A new soft-rank fully supervised robust regression model. This convex model is able to deal with noise, outliers and missing data in the input variables. (4) A new soft-rank model for weakly supervised image classification and localization. Unlike existing multiple-instance approaches for this problem, our model is convex.

This operations research thesis should be situated in the field of the power generation industry. The general objective of this work is to efficiently solve the Generalized Unit Commitment (GUC) problem by means of specialized software. The GUC problem generalizes the Unit Commitment (UC) problem by simultane-ously solving the associated Optimal Power Flow (OPF) problem. There are many approaches to solve the UC and OPF problems separately, but approaches to solve them jointly, i.e. to solve the GUC problem, are quite scarce. One of these GUC solving approaches is due to professors Batut and Renaud, whose methodology has been taken as a starting point for the methodology presented herein.This thesis report is structured as follows. Chapter 1 describes the state of the art of the UC and GUC problems. The formulation of the classical short-term power planning problems related to the GUC problem, namely the economic dispatching problem, the OPF problem, and the UC problem, are reviewed. Special attention is paid to the UC literature and to the traditional methods for solving the UC problem. In chapter 2 we extend the OPF model developed by professors Heredia and Nabona to obtain our GUC model. The variables used and the modelling of the thermal, hydraulic and transmission systems are introduced, as is the objective function. Chapter 3 deals with the Variable Duplication (VD) method, which is used to decompose the GUC problem as an alternative to the Classical Lagrangian Relaxation (CLR) method. Furthermore, in chapter 3 dual bounds provided by the VDmethod or by the CLR methods are theoretically compared.Throughout chapters 4, 5, and 6 our solution methodology, the Radar Multiplier (RM) method, is designed and tested. Three independent matters are studied: first, the auxiliary problem principle method, used by Batut and Renaud to treat the inseparable augmented Lagrangian, is compared with the block coordinate descent method from both theoretical and practical points of view. Second, the Radar Sub- gradient (RS) method, a new Lagrange multiplier updating method, is proposed and computationally compared with the classical subgradient method. And third, we study the local character of the optimizers computed by the Augmented Lagrangian Relaxation (ALR) method when solving the GUC problem. A heuristic to improve the local ALR optimizers is designed and tested.Chapter 7 is devoted to our computational implementation of the RM method, the MACH code. First, the design of MACH is reviewed brie y and then its performance is tested by solving real-life large-scale UC and GUC instances. Solutions computed using our VD formulation of the GUC problem are partially primal feasible since they do not necessarily fulfill the spinning reserve constraints. In chapter 8 we study how to modify this GUC formulation with the aim of obtaining full primal feasible solutions. A successful test based on a simple UC problem is reported. The conclusions, contributions of the thesis, and proposed further research can be found in chapter 9.

Ce mémoire de thèse à été réalisée dans le cadre d'une collaboration scientifique avec "La Manufacture Française des Pneumatiques Michelin". Il porte sur l'analyse mathématique et numérique de la convergence et de la stabilité de formulations mixtes ou hybrides de problèmes d'optimisation sous contrainte avec la méthode des multiplicateurs de Lagrange et dans le cadre de la méthode éléments finis étendus (XFEM). Tout d'abord, nous essayons de démontrer la stabilité de la discrétisation X-FEM pour le problème d'élasticité linéaire incompressible en statique. Le deuxième axe, qui représente le contenu principal de la thèse est dédié à l'étude de certaines méthodes de multiplicateur de Lagrange stabilisées. La particularité de ces méthodes est que la stabilité du multiplicateur est assurée par l'ajout de termes supplémentaires dans la formulation faible. Dans ce contexte, nous commençons par l'étude de la méthode de stabilisation de Barbosa-Hughes appliquée au problème de contact unilatéral sans frottement avec XFEM cut-off. Ensuite, nous construisons une nouvelle méthode basée sur des techniques de projections locales pour stabiliser un problème de Dirichlet dans le cadre de X-FEM et une approche de type domaine fictif. Nous faisons aussi une étude comparative entre la stabilisation avec la technique de projection locale et la stabilisation de Barbosa-Hughes. Enfin, nous appliquons cette nouvelle méthode de stabilisation aux problèmes de contact unilatéral en élastostatique avec frottement de Tresca dans le cadre de X-FEM. / This Ph.D. thesis was done in collaboration with "La Manufacture Française des Pneumatiques Michelin". It concerns the mathematical and numerical analysis of convergence and stability of mixed or hybrid formulation of constrained optimization problem with Lagrange multiplier method in the framework of the eXtended Finite Element Method (XFEM). First we try to prove the stability of the X-FEM discretization for incompressible elastostatic problem by ensured a LBB condition. The second axis, which present the main content of the thesis, is dedicated to the use of some stabilized Lagrange multiplier methods. The particularity of these stabilized methods is that the stability of the multiplier is provided by adding supplementary terms in the weak formulation. In this context, we study the Barbosa-Hughes stabilization technique applied to the frictionless unilateral contact problem with XFEM-cut-off. Then we present a new consistent method based on local projections for the stabilization of a Dirichlet condition in the framework of extended finite element method with a fictitious domain approach. Moreover we make comparative study between the local projection stabilization and the Barbosa-Hughes stabilization. Finally we use the local projection stabilization to approximate the two-dimensional linear elastostatics unilateral contact problem with Tresca frictional in the framework of the eXtended Finite Element Method X-FEM.

La thèse est portée essentiellement sur la stabilisation et la contrôlabilité de deux équations des ondes moyennant un seul contrôle agissant sur le bord du domaine. Dans le cas du contrôle dynamique, le contrôle est introduit dans le système par une équation différentielle agissant sur le bord. C'est en effet un système hybride. Le contrôle peut être aussi applique directement sur le bord d'une équation, c'est le cas du contrôle indirecte mais non borne. La nature du système ainsi coupledépend du couplage des équations, et ceci donne divers résultats par la stabilisation (exponentielle et polynomiale) et la contrôlabilité exacte (espace contrôlable). Des nouvelles inégalités d'énergie permettent de mettre en oeuvre la Méthode fréquentielle et la Méthode d'Unicité de Hilbert. / This thesis is concerned with the stabilization and the exact controllability of two wave equations by means of only one control acting on the boundary of the domain. In the case of dynamic control, the control is introduced into the system by differential equation acting on the boundary. It is indeed a hybrid system. The control can be also applied directly on the boundary of one of the equations. In this case, the control is indirect but unbounded. The behavior of the obtained system depends on theways of coupling. Various results are established for the stabilization (exponential or polynomial) and the exact controllability (controllable space of initial data). A new inequality of energy allows to apply the Frequency Method and the Hilbert Uniqueness Method.

Cette thèse est consacrée à l’étude de la stabilisation de certains systèmes distribués avec contrôle frontière de type dynamique. Nous considérons, d’abord, la stabilisation de l’équation de la poutre de Rayleigh avec un seul contrôle frontière dynamique moment ou force. Nous montrons que le système n’est pas uniformément (autrement dit exponentiellement) stable; mais par une méthode spectrale, nous établissons le taux polynomial optimal de décroissance de l’énergie du système. Ensuite, nous étudions la stabilisation indirecte de l’équation des ondes avec un amortissement frontière de type dynamique fractionnel. Nous montrons que le taux de décroissance de l’énergie dépend de la nature géométrique du domaine. En utilisant la méthode fréquentielle et une méthode spectrale, nous montrons la non stabilité exponentielle et nous établissons, plusieurs résultats de stabilité polynomiale. Enfin, nous considérons l’approximation de l’équation des ondes mono-dimensionnelle avec un seul amortissement frontière de type dynamique par un schéma de différence finie. Par une méthode spectrale, nous montrons que l’énergie discrétisée ne décroit pas uniformément (par rapport au pas du maillage) polynomialement vers zéro comme l’énergie du système continu. Nous introduisons, alors, un terme de viscosité numérique et nous montrons la décroissance polynomiale uniforme de l’énergie de notre schéma discret avec ce terme de viscosité. / This thesis is devoted to the study of the stabilization of some distributed systems with dynamic boundary control. First, we consider the stabilization of the Rayleigh beam equation with only one dynamic boundary control moment or force. We show that the system is not uniformly (exponentially) stable. However, using a spectral method, we establish the optimal polynomial decay rate of the energy of the system. Next, we study the indirect stability of the wave equation with a fractional dynamic boundary control. We show that the decay rate of the energy depends on the nature of the geometry of the domain. Using a frequency approach and a spectral method, we show the non exponential stability of the system and we establish, different polynomial stability results. Finally, we consider the finite difference space discretization of the 1-d wave equation with dynamic boundary control. First, using a spectral approach, we show that the polynomial decay of the discretized energy is not uniform with respect to the mesh size, as the energy of the continuous system. Next, we introduce a viscosity term and we establish the uniform (with respect to the mesh size) polynomial energy decay of our discrete scheme.

科学研究費補助金 研究種目:基盤研究(C) 課題番号:10650238 研究代表者:安田 仁彦 研究期間:1998-1999年度