Approche de commande de systèmes à événements discrets via des techniques de comparaison stochastique


Book Description

Ce travail porte sur la commande de systèmes à événements discrets via des techniques de comparaison stochastique. Les systèmes à événements discrets sont des systèmes dynamiques qui évoluent de façon discontinue. On s’intéresse en particulier à ceux de ces systèmes qui peuvent être représentés par des itérés de fonctions linéaires aléatoires (chaînes de Markov), ou par des itérés de fonctions linéaires dans l'algèbre des dioïdes (systèmes max-plus linéaires). D'abord, nous nous sommes intéressés à l'évaluation de mesures quantitatives du comportement en régime transitoire de chaînes de Markov. Pour les processus issus du monde réel, on se heurte au problème de l’explosion combinatoire du temps de calculs et de la taille de l'espace d’états. Pour y remédier, des techniques d'agrégation bornantes basées sur la comparaison de chaînes de Markov de dimensions différentes, dont le but principal est le calcul de bornes, sont proposées. Ces méthodes de comparaison sont inspirées par la théorie des ordres stochastiques et sont formulées sous la forme de plusieurs critères, à savoir un critère géométrique basé sur l’inclusion de polyèdres, un critère d’invariance positive d’ensembles, et un critère algébrique basé sur le Lemme de Haar donné en 1918. Des concepts similaires peuvent être énoncés pour des systèmes max-plus linéaires. Comme pour les chaînes de Markov, des techniques de comparaison de systèmes max-plus linéaires sont proposées, dont les deux objectifs principaux sont la simplification de modèles et le contrôle via l’invariance positive d’ensembles. Enfin, nous explorons la propriété d’invariance positive d’ensemble max-plus linéaires et en particulier des ensembles max-plus ellipsoïdaux. Ces ensembles ne sont autres que des ensembles polyédriques dans l’algèbre linéaire habituelle. Les caractérisations d’invariance positive de telles classes de systèmes max-plus linéaires sont formulées sous la forme d’inclusion de polyèdres dans l’algèbre linéaire. On en déduit des conditions d’existence et de calcul de lois de commande par retour d’état linéaire statique, basé sur la gamma-algorithme, pour des systèmes dynamiques.




Approche probabiliste pour la commande orientée évènement des systèmes stochastiques à commutation


Book Description

Les systèmes hybrides sont des systèmes dynamiques, caractérisés par un comportementdual, une interaction entre une partie discrète et une partie continue de fonctionnement.Dans le centre de notre travail se trouve une classe particulière de systèmeshybrides, plus spécifiquement les systèmes stochastiques à commutation que nous modélisonsà l'aide des Chaînes de Markov en temps continu et des équations différentielles.Le comportement aléatoire de ce type de système nécessite une commande spécialequi s'adapte aux événements arbitraires qui peuvent changer complètement l'évolutiondu système. Nous avons choisi une politique de contrôle basée sur les événements quiest déclenchée seulement quand il est nécessaire (sur un événement incontrôlable - parexemple un seuil qui est atteint), et jusqu'à ce que certaines conditions de fonctionnementsont remplies (le système revient dans l'état normal).Notre approche vise le développement d'un modèle probabiliste qui permet de calculerun critère de performance (en occurrence l'énergie du système) pour la politiquede contrôle choisie. Nous proposons d'abord une méthode de simulation à événementsdiscrets pour le système stochastique à commutation commandé, qui nous donne la possibilitéde réaliser une optimisation directe de la commande appliquée sur le système etaussi de valider les modèles analytiques que nous avons construit pour l'application ducontrôle.Un modèle analytique pour déterminer l'énergie consommée par le système a étéconçu en utilisant les probabilités de sortie de la région de contrôle, respectivement lestemps de séjour dans la chaîne de Markov avant et après avoir atteint les limites decontrôle. La dernière partie du travail présente la comparaison des résultats obtenusentre la méthode analytique et la simulation.




Statistical Inference for Ergodic Diffusion Processes


Book Description

The first book in inference for stochastic processes from a statistical, rather than a probabilistic, perspective. It provides a systematic exposition of theoretical results from over ten years of mathematical literature and presents, for the first time in book form, many new techniques and approaches.




Semiparametric Theory and Missing Data


Book Description

This book summarizes current knowledge regarding the theory of estimation for semiparametric models with missing data, in an organized and comprehensive manner. It starts with the study of semiparametric methods when there are no missing data. The description of the theory of estimation for semiparametric models is both rigorous and intuitive, relying on geometric ideas to reinforce the intuition and understanding of the theory. These methods are then applied to problems with missing, censored, and coarsened data with the goal of deriving estimators that are as robust and efficient as possible.




Quantum Mechanics, Volume 3


Book Description

This new, third volume of Cohen-Tannoudji's groundbreaking textbook covers advanced topics of quantum mechanics such as uncorrelated and correlated identical particles, the quantum theory of the electromagnetic field, absorption, emission and scattering of photons by atoms, and quantum entanglement. Written in a didactically unrivalled manner, the textbook explains the fundamental concepts in seven chapters which are elaborated in accompanying complements that provide more detailed discussions, examples and applications. * Completing the success story: the third and final volume of the quantum mechanics textbook written by 1997 Nobel laureate Claude Cohen-Tannoudji and his colleagues Bernard Diu and Franck Laloë * As easily comprehensible as possible: all steps of the physical background and its mathematical representation are spelled out explicitly * Comprehensive: in addition to the fundamentals themselves, the books comes with a wealth of elaborately explained examples and applications Claude Cohen-Tannoudji was a researcher at the Kastler-Brossel laboratory of the Ecole Normale Supérieure in Paris where he also studied and received his PhD in 1962. In 1973 he became Professor of atomic and molecular physics at the Collège des France. His main research interests were optical pumping, quantum optics and atom-photon interactions. In 1997, Claude Cohen-Tannoudji, together with Steven Chu and William D. Phillips, was awarded the Nobel Prize in Physics for his research on laser cooling and trapping of neutral atoms. Bernard Diu was Professor at the Denis Diderot University (Paris VII). He was engaged in research at the Laboratory of Theoretical Physics and High Energy where his focus was on strong interactions physics and statistical mechanics. Franck Laloë was a researcher at the Kastler-Brossel laboratory of the Ecole Normale Supérieure in Paris. His first assignment was with the University of Paris VI before he was appointed to the CNRS, the French National Research Center. His research was focused on optical pumping, statistical mechanics of quantum gases, musical acoustics and the foundations of quantum mechanics.




Flight Vehicle System Identification


Book Description

This valuable volume offers a systematic approach to flight vehicle system identification and exhaustively covers the time domain methodology. It addresses in detail the theoretical and practical aspects of various parameter estimation methods, including those in the stochastic framework and focusing on nonlinear models, cost functions, optimization methods, and residual analysis. A pragmatic and balanced account of pros and cons in each case is provided. The book also presents data gathering and model validation, and covers both large-scale systems and high-fidelity modeling. Real world problems dealing with a variety of flight vehicle applications are addressed and solutions are provided. Examples encompass such problems as estimation of aerodynamics, stability, and control derivatives from flight data, flight path reconstruction, nonlinearities in control surface effectiveness, stall hysteresis, unstable aircraft, and other critical considerations.




Predicting Structured Data


Book Description

State-of-the-art algorithms and theory in a novel domain of machine learning, prediction when the output has structure.







Effective Actuarial Methods


Book Description

During the last two decades actuarial research has developed in a more applied direction. Although the original risk models generally served as convenient and sometimes tractable mathematical examples of general probabilistic and/or statistical theories, nowadays models and techniques are encountered that can be considered to be typically actuarial. Examples include ordering of risks by dangerousness, credibility theory and techniques based on IBNR models. Not only does this book present the underlying mathematics of these subjects, but it also deals with the practical application of the techniques. In order to provide results based on real insurance portfolios, use is made of three software packages, namely SLIC performing stop-loss insurance calculations for individual and collective risk models, CRAC dealing with actuarial applications of credibility theory, and LORE giving IBNR-based estimates for loss reserves. Worked-out examples illustrate the theoretical results. This book is intended for use in preparing university actuarial exams, and contains many exercises with varying levels of complexity. It is valuable as a textbook for students in actuarial sciences during their last year of study. Due to the emphasis on applications and because of the worked-out examples on real portfolio data, it is also useful for practising actuaries to guide them in interpreting their own results.




An Introduction to Computational Learning Theory


Book Description

Emphasizing issues of computational efficiency, Michael Kearns and Umesh Vazirani introduce a number of central topics in computational learning theory for researchers and students in artificial intelligence, neural networks, theoretical computer science, and statistics. Emphasizing issues of computational efficiency, Michael Kearns and Umesh Vazirani introduce a number of central topics in computational learning theory for researchers and students in artificial intelligence, neural networks, theoretical computer science, and statistics. Computational learning theory is a new and rapidly expanding area of research that examines formal models of induction with the goals of discovering the common methods underlying efficient learning algorithms and identifying the computational impediments to learning. Each topic in the book has been chosen to elucidate a general principle, which is explored in a precise formal setting. Intuition has been emphasized in the presentation to make the material accessible to the nontheoretician while still providing precise arguments for the specialist. This balance is the result of new proofs of established theorems, and new presentations of the standard proofs. The topics covered include the motivation, definitions, and fundamental results, both positive and negative, for the widely studied L. G. Valiant model of Probably Approximately Correct Learning; Occam's Razor, which formalizes a relationship between learning and data compression; the Vapnik-Chervonenkis dimension; the equivalence of weak and strong learning; efficient learning in the presence of noise by the method of statistical queries; relationships between learning and cryptography, and the resulting computational limitations on efficient learning; reducibility between learning problems; and algorithms for learning finite automata from active experimentation.