La simulation numérique est indissociable du calcul haute performance. Ces vingt dernières années,l'informatique a connu l'émergence d'architectures parallèles multi-niveaux. Exploiter efficacement lapuissance de calcul de ces machines peut s'avérer être une tâche délicate et requérir une expertise à la foistechnologique sur des notions avancées de parallélisme ainsi que scientifique de part la nature même desproblèmes traités.Le travail de cette thèse est pluri-disciplinaire s'appuyant sur la conception d'une librairie de calculparallèle réutilisable pour la résolution des équations Hamilton-Jacobi-Bellman. Ces équations peuventse retrouver dans des domaines diverses et variés tels qu'en biomédical, géophysique, ou encore robotiqueen l'occurence sur les applications de planification de mouvement et de reconstruction de formestri-dimensionnelles à partir d'images bi-dimensionnelles. Nous montrons que les principaux algorithmesnumériques amenant a résoudre ces équations telles que les méthodes de type fast marching, ne sont pasappropriés pour être efficaces dans un contexte parallèle. Nous proposons la méthode buffered fast iterativequi permet d'obtenir une scalabilité parallèle non obtenue jusqu'alors. Un des points sensibles relevésdans cette thèse est de parvenir à trouver une recette de compromis entre abstraction, performance etmaintenabilité afin de garantir non seulement une réutilisabilitédans le sens classique du domaine de génielogiciel mais également en terme de réutilisabilité séquentielle/parallèle / Numerical simulation is strongly bound with high performance computing. Programming scientificsoftwares requires at the same time good knowledge on the mathematical numerical models and alsoon the techniques to make them efficient on today's computers. Indeed, these last twenty years, wehave experienced the rising of multi-level parallel architectures. The work in this thesis dissertation ismultidisciplinary by designing a reusable parallel numerical library for solving Hamilton-Jacobi-Bellmanequations. Such equations are involved in various fields such as in biomedical, geophysics or robotics. Inparticular, we will show interests in path planning and shape from shading applications. We show thatthe methods to solve these equations such as the widely used fast marching method, are not designedto be used effciently in a parallel context. We propose a buffered fast iterative method which givesan interesting parallel scalability. This dissertation takes interest in the challenge to find compromisesbetween abstraction, performance and maintainability in order to combine both software reusability andalso sequential/parallel reusability. We propose code abstraction allowing algorithmic and data genericitywhile trying to keep a maintainable and performant code potentially parallelizable

Submitted by Maria Aparecida (cidazen@gmail.com) on 2017-09-04T13:42:58Z No. of bitstreams: 1 Watson Robert.pdf: 2699368 bytes, checksum: cf204eec3df50b251f4adbbbd380ffd0 (MD5) / Made available in DSpace on 2017-09-04T13:42:58Z (GMT). No. of bitstreams: 1 Watson Robert.pdf: 2699368 bytes, checksum: cf204eec3df50b251f4adbbbd380ffd0 (MD5) Previous issue date: 2014-08-21 / Due to the demand for high-performance equipments and the rising cost of energy, the industrial sector is developing equipments to attend minimization of the theirs operational costs. The implementation of these requirements generate a demand for projects and implementations of high-performance control systems. The optimal control theory is an alternative to solve this problem, because in its design considers the normative specifications of the system design, as well as those that are related to the operational costs. Motivated by these perspectives, it is presented the study of methods and the development of algorithms to the approximated solution of the Equation Hamilton-Jacobi-Bellman, in the form of discrete Riccati equation, model free and dependent of the dynamic system. The proposed solutions are developed in the context of adaptive dynamic programming that are based on the methods for online design of optimal control systems, Discrete Linear Quadratic Regulator type. The proposed approach is evaluated in multivariable models of the dynamic systems to evaluate the perspectives of the optimal control law for online implementations. / Devido a demanda por equipamentos de alto desempenho e o custo crescente da energia, o setor industrial desenvolve equipamentos que atendem a minimização dos seus custos operacionais. A implantação destas exigências geram uma demanda por projetos e implementações de sistemas de controle de alto desempenho. A teoria de controle ótimo é uma alternativa para solucionar este problema, porque considera no seu projeto as especificações normativas de projeto do sistema, como também as relativas aos seus custos operacionais. Motivado por estas perspectivas, apresenta-se o estudo de métodos e o desenvolvimento de algoritmos para solução aproximada da Equação Hamilton-Jacobi-Bellman, do tipo Equação Discreta de Riccati, livre e dependente de modelo do sistema dinâmico. As soluções propostas são desenvolvidas no contexto de programação dinâmica adaptativa (ADP) que baseiam-se nos métodos para o projeto on-line de Controladores Ótimos, do tipo Regulador Linear Quadrático Discreto. A abordagem proposta é avaliada em modelos de sistemas dinâmicos multivariáveis, tendo em vista a implementação on-line de leis de controle ótimo.

Cette thèse est composée de quatre chapitres. Le premier chapitre traite de l'évaluation de produits financiers dans un modèle comportant un saut pour l'actif risque. Ce saut représente la faillite de l'entreprise correspondante. On étudie alors l'évaluation des prix d'options par indifférence d'utilité dans un cadre d'utilité exponentielle. Par des techniques de programmation dynamique on montre que le prix d'un Bond est solution d'une équation différentielle et le prix d'options dépendantes de l'actif est solution d'une équation aux dérives partielles d'Hamilton-Jacobi-Bellman. Le saut dans la dynamique de l'actif risque induit des différences avec le modèle de Merton que nous tentons de quantifier. Le second chapitre traite d'un marché comportant des sauts : les paris sur le football. Nous rappelons les différentes familles de modèles pour un match de football et introduisons un modèle complet permettant d'évaluer les prix des différents produits apparus sur ce marché ces dix dernières années. La complexité de ce modèle nous amène à étudier un modèle simplifié dont nous étudions les implications et calculons les prix obtenus que l'on compare à la réalité. On remarque que la calibration implicite obtenue génère de très bons résultats en produisant des prix très proches de la réalité. Le troisième chapitre développe le problème de fixation des prix par un teneur de marche monopolistique dans le marché des paris binaires. Ce travail est un prolongement direct au problème introduit par Levitt [Lev04]. Nous généralisons en effet son travail aux cas des paris européens et proposons une méthode pour estimer la méthode de cotation utilisée par le book-maker. Nous montrons que deux hypothèses inextricables peuvent expliquer cette fixation des prix. D'une part, l'incertitude du public sur la vraie valeur ainsi que le caractère extrêmement risque-averse du bookmaker. Le quatrième chapitre prolonge quant à lui cette approche au cas de produits financiers non binaires. Nous examinons différents modèles d'offre et de demande et en déduisons, par des techniques de programmation dynamique, des équations aux dérivées partielles dictant la formation des prix d'achat et de vente. Nous montrons finalement que l'écart entre prix d'achat et prix de vente ne dépend pas de la position du teneur de marche dans l'actif considère. Cependant le prix moyen dépend lui fortement de la quantité détenue par le teneur de marche. Une approche simplifiée est finalement proposée dans le cas multidimensionnel / This thesis is composed of four chapters. The first one deals with the pricing of financial products in a single jump model for the risky asset. This jump represents the bankrupcy of the quoted firm. We study the pricing of derivatives in the context of indifference of utility with an exponential utility. By means of dynamic programming we show that the bond price is solution of an ordinary differential equation and that stock price dependent options are solutions of an equation with partial derivatives of Hamilton-Jacobi-Bellman type generalizing the Black-Scholes one. We then try to quantify differences in the price obtained here and the one from Merton model without jump. The second chapter deals with a specific jump market : the soccer betting market. We recall the different model families for a soccer match and introduce some full model which allows to price the products recently born in this market in last ten years. Nevertheless the model complexity leads us to study a simplified model introduced by Dixon and Robinson from which we are able to derive closed formulas and simulate prices that we compare to market prices. We remark that implicit calibration gives pretty goof fit of market data. Third chapter developps the approach of Levitt [Lev04] on price formation in binary betting market held by a monopolistic market-maker operating in a one time step trading. We generalize Levitt results with european format of betting. We show that prices are distorded on the pressure of demand and offer, that phenomena introducing a market probability that allows to price products under this new measure. We identify some best model for demand and offer and market maker strategy and show that probability change is obvious in case of imperfect information about the value of the product. Fourth chapter generalizes this approach to the case of general payoffs and continuous time. The task is more complex and we just derive partial derivative equations from dynamic programming that enable us to give the bid-ask prices of the product traded by the market-maker. One result is that, in most models, bid-ask spread does not depend on the inventory held by the dealer whereas mid-quote price strongly reflects the unbalance of the dealer

In this thesis we study the optimal stochastic control problem of the drift of a Lévy process. We show that, for a broad class of Lévy processes, the partial integro-differential Hamilton-Jacobi-Bellman equation for the value function admits classical solutions and that control policies exist in feedback form. We then explore the class of Lévy processes that satisfy the requirements of the theorem, and find connections between the uniform integrability requirement and the notions of the score function and Fisher information from information theory. Finally we present three different numerical implementations of the control problem: a traditional dynamic programming approach, and two iterative approaches, one based on a finite difference scheme and the other on the Fourier transform. / text

>Magister Scientiae - MSc / Bank asset management mainly involves profit maximization through invest- ment in loans giving high returns on loans, investment in securities for reducing risk and providing liquidity needs. In particular, commercial banks grant loans to creditors who pay high interest rates and are not likely to default on their loans. Furthermore, the banks purchase securities with high returns and low risk. In addition, the banks attempt to lower risk by diversifying their asset portfolio. The main categories of assets held by banks are loans, treasuries (bonds issued by the national treasury), reserves and intangible assets. In this mini-thesis, we solve an optimal asset allocation problem in banking under the mean-variance frame work. The dynamics of the different assets are modelled as geometric Brownian motions, and our optimization problem is of the mean- variance type. We assume the Basel II regulations on banking supervision. In this contribution, the bank funds are invested into loans and treasuries with the main objective being to obtain an optimal return on the bank asset port- folio given a certain risk level. There are two main approaches to portfolio optimization, which are the so called martingale method and Hamilton Jacobi Bellman method. We shall follow the latter. As is common in portfolio op- timization problems, we obtain an explicit solution for the value function in the Hamilton Jacobi Bellman equation. Our approach to the portfolio prob- lem is similar to the presentation in the paper [Hojgaard, B., Vigna, E., 2007. Mean-variance portfolio selection and efficient frontier for defined contribution pension schemes. ISSN 1399-2503. On-line version ISSN 1601-7811]. We pro- vide much more detail and we make the application to banking. We illustrate our findings by way of numerical simulations.

Optimal control problems can in many cases become complicated and difficult to solve. One particular class of difficult control problems to solve are singular control problems. Standard methods for solving optimal control are discussed showing why those methods are difficult to apply to singular control problems. Then standard methods for solving singular control problems are discussed including why the standard methods can be difficult and often impossible to apply without having to resort to numerical techniques. Finally, an alternative method to solving a class of singular optimal control problems is given for a specific class of problems.

Asymptotic stability is a key notion of system stability for controlled dynamical systems as it guarantees that the system trajectories are bounded in a neighborhood of a given isolated equilibrium point and converge to this equilibrium over the infinite horizon. In some applications, however, asymptotic stability is not the appropriate notion of stability. For example, for systems with a continuum of equilibria, every neighborhood of an equilibrium contains another equilibrium and a nonisolated equilibrium cannot be asymptotically stable. Alternatively, in stabilization of spacecraft dynamics via gimballed gyroscopes, it is desirable to find state- and output-feedback control laws that guarantee partial-state stability of the closed-loop system, that is, stability with respect to part of the system state. Furthermore, we may additionally require finite-time stability of the closed-loop system, that is, convergence of the system's trajectories to a Lyapunov stable equilibrium in finite time. The Hamilton-Jacobi-Bellman optimal control framework provides necessary and sufficient conditions for the existence of state-feedback controllers that minimize a given performance measure and guarantee asymptotic stability of the closed-loop system. In this research, we provide extensions of the Hamilton-Jacobi-Bellman optimal control theory to develop state-feedback control laws that minimize nonlinear-nonquadratic performance criteria and guarantee semistability, partial-state stability, finite-time stability, and finite-time partial state stability of the closed-loop system.

L'objectif de cette thèse est de développer la théorie du contrôle stochastique et ses applications en dynamique des populations. D'un point de vue théorique, nous présentons l'étude de problèmes de contrôle stochastique à horizon fini sur des processus de diffusion, de branchement non linéaire et de branchement-diffusion. Dans chacun des cas, nous raisonnons par la méthode de la programmation dynamique en veillant à démontrer soigneusement un argument de conditionnement analogue à la propriété de Markov forte pour les processus contrôlés. Le principe de la programmation dynamique nous permet alors de prouver que la fonction valeur est solution (régulière ou de viscosité) de l'équation de Hamilton-Jacobi-Bellman correspondante. Dans le cas régulier, nous identifions également un contrôle optimal markovien par un théorème de vérification. Du point de vue des applications, nous nous intéressons à la modélisation mathématique du cancer et de ses stratégies thérapeutiques. Plus précisément, nous construisons un modèle hybride de croissance de tumeur qui rend compte du rôle fondamental de l'acidité dans l'évolution de la maladie. Les cibles de la thérapie apparaissent explicitement comme paramètres du modèle afin de pouvoir l'utiliser comme support d'évaluation de stratégies thérapeutiques.

The existence of pathwise stationary solutions of this stochastic partial differential equation (SPDE, for abbreviation) is demonstrated. In Part II, a connection between certain kind of state constrained controlled Forward-Backward Stochastic Differential Equations (FBSDEs) and Hamilton-Jacobi-Bellman equations (HJB equations) are demonstrated. The special case provides a probabilistic representation of some geometric flows, including the mean curvature flows. Part II includes also a probabilistic proof of the finite time existence of the mean curvature flows.

In this dissertation, certain problems of stochastic optimal control and relevant analysis of random vibrations are considered. Dynamic Programming approach is used to find an optimal control law for a linear single-degree-of-freedom system subjected to Gaussian white-noise excitation. To minimize a system's mean response energy, a bounded in magnitude control force is applied. This approach reduces the problem of finding the optimal control law to a problem of finding a solution to the Hamilton-Jacobi-Bellman (HJB) partial differential equation. A solution to this partial differential equation (PDE) is obtained by developed 'hybrid' solution method. The application of bounded in magnitude control law will always introduce a certain type of nonlinearity into the system's stochastic equation of motion. These systems may be analyzed by the Energy Balance method, which introduced and developed in this dissertation. Comparison of analytical results obtained by the Energy Balance method and by stochastic averaging method with numerical results is provided. The comparison of results indicates that the Energy Balance method is more accurate than the well-known stochastic averaging method.