Ce travail présente quelques résultats sur les systèmes de particules en interaction pour l'interprétation probabiliste des équations aux dérivées partielles, avec des applications à des questions de dynamique moléculaire et de chimie quantique. On présente notamment une méthode particulaire permettant d'analyser le processus de la force biaisante adaptative, utilisé en dynamique moléculaire pour le calcul de différences d'énergies libres. On étudie également la sensibilité de dynamiques stochastiques par rapport à un paramètre, en vue du calcul des forces dans l'approximation de Born-Oppenheimer pour rechercher l'état quantique fondamental de molécules. Enfin, on présente un schéma numérique basé sur un système de particules pour résoudre des lois de conservation scalaires, avec un terme de diffusion anormale se traduisant par une dynamique de sauts sur les particules / This work presents some results on stochastically interacting particle systems and probabilistic interpretations of partial differential equations with applications to molecular dynamics and quantum chemistry. We present a particle method allowing to analyze the adaptive biasing force process, used in molecular dynamics for the computation of free energy differences. We also study the sensitivity of stochastic dynamics with respect to some parameter, aiming at the computation of forces in the Born-Oppenheimer approximation for determining the fundamental quantum state of molecules. Finally, we present a numerical scheme based on a particle system for the resolution of scalar conservation laws with an anomalous diffusion term, corresponding to a jump dynamics on the particles

In this thesis, we present a new existence result for Lévy-type processes. Lévy-type processes behave locally like a Lévy process, but the Lévy triplet may depend on the current position of the process. They can be characterized by their so-called symbol; this is the analogue of the characteristic exponent in the Lévy case. Using a parametrix construction, we prove the existence of Lévy-type processes with a given symbol under weak regularity assumptions on the regularity of the symbol. Applications range from existence results for stable-like processes and mixed processes to uniqueness results for Lévy-driven stochastic differential equations. Moreover, we discuss sufficient conditions for the existence of moments of Lévy-type processes and derive estimates for fractional moments.

Cette thèse étudie la gestion et la couverture du risque en s’appuyant sur la Value-at-Risk (VaR) et la Value-at-Risk Conditionnelle (CVaR), comme mesures de risque. La première partie propose un modèle d’évolution de prix que nous confrontons à des données réelles issues de la bourse de Paris (Euronext PARIS). Notre modèle prend en compte les probabilités d’occurrence des pertes extrêmes et les changements de régimes observés sur les données. Notre approche consiste à détecter les différentes périodes de chaque régime par la construction d’une chaîne de Markov cachée et à estimer la queue de distribution de chaque régime par des lois puissances. Nous montrons empiriquement que ces dernières sont plus adaptées que les lois normales et les lois stables. L’estimation de la VaR est validée par plusieurs backtests et comparée aux résultats d’autres modèles classiques sur une base de 56 actifs boursiers. Dans la deuxième partie, nous supposons que les prix boursiers sont modélisés par des exponentielles de processus de Lévy. Dans un premier temps, nous développons une méthode numérique pour le calcul de la VaR et la CVaR cumulatives. Ce problème est résolu en utilisant la formalisation de Rockafellar et Uryasev, que nous évaluons numériquement par inversion de Fourier. Dans un deuxième temps, nous nous intéressons à la minimisation du risque de couverture des options européennes, sous une contrainte budgétaire sur le capital initial. En mesurant ce risque par la CVaR, nous établissons une équivalence entre ce problème et un problème de type Neyman-Pearson, pour lequel nous proposons une approximation numérique s’appuyant sur la relaxation de la contrainte / This thesis studies the risk management and hedging, based on the Value-at-Risk (VaR) and the Conditional Value-at-Risk (CVaR) as risk measures. The first part offers a stocks return model that we test in real data from NSYE Euronext. Our model takes into account the probability of occurrence of extreme losses and the regime switching observed in the data. Our approach is to detect the different periods of each regime by constructing a hidden Markov chain and estimate the tail of each regime distribution by power laws. We empirically show that powers laws are more suitable than Gaussian law and stable laws. The estimated VaR is validated by several backtests and compared to other conventional models results on a basis of 56 stock market assets. In the second part, we assume that stock prices are modeled by exponentials of a Lévy process. First, we develop a numerical method to compute the cumulative VaR and CVaR. This problem is solved by using the formalization of Rockafellar and Uryasev, which we numerically evaluate by Fourier inversion techniques. Secondly, we are interested in minimizing the hedging risk of European options under a budget constraint on the initial capital. By measuring this risk by CVaR, we establish an equivalence between this problem and a problem of Neyman-Pearson type, for which we propose a numerical approximation based on the constraint relaxation

Cette thèse est consacrée à l’étude des équations aux dérivées partielles stochastiques de type parabolique. Dans la première partie nous démontrons de nouveaux résultats concernant l’existence et l’unicité de solutions variationnelles globales et locales à des problèmes avec des conditions aux bords de type Neumann pour une classe d’équations aux dérivées partielles stochastiques non-autonomes. Les équations que nous considérons sont définies sur des domaines non bornés de l’espace euclidien qui satisfont à certaines conditions géométriques, et sont dirigées par un bruit multiplicatif dérivé d’un processus de Wiener fractionnaire infini-dimensionnel caractérisé par une suite de paramètres de Hurst H = (Hi) i ∈ N+ ⊂ (1/2,1). Ces paramètres sont en fait soumis à d’autres contraintes intimement liées à la nature de la non-linéarité dans le terme stochastique des équations, et au choix des espaces fonctionnels dans lesquels le problème à résoudre est bien posé. Notre méthode de preuve repose essentiellement sur des arguments d’injections compactes. Dans la seconde partie, nous étudions la possibilité de l’explosion de solutions d’une classe d’équations aux dérivées partielles stochastiques semi-linéaire avec des conditions aux bords de type Dirichlet, perturbées par un mélange d’un mouvement brownien et d’un mouvement brownien fractionnaire et dirigées par une classe d’opérateurs différentiels non autonomes contenant des processus de diffusions et des processus de Lévy. Notre but est de comprendre l’influence de la partie stochastique et de l’opérateur différentiel sur le comportement d’explosion des solutions. En particulier, nous donnons des expressions explicites pour des bornes inférieures et supérieures du temps de l’explosion de la solution, et des conditions suffisantes pour l’existence d’une solution globale positive. Nous estimons également la probabilité d’une explosion en temps fini et la loi d’une borne supérieur du temps d’explosion de la solution / This thesis is concerned with stochastic partial differential equations of parabolic type. In the first part we prove new results regarding the existence and the uniqueness of global and local variational solutions to a Neumann initial-boundary value problem for a class of non-autonomous stochastic parabolic partial differential equations. The equations we consider are defined on unbounded open domains in Euclidean space satisfying certain geometric conditions, and are driven by a multiplicative noise derived from an infinite-dimensional fractional Wiener process characterized by a sequence of Hurst parameters H = (Hi) i ∈ N+ ⊂ (1/2,1). These parameters are in fact subject to further constraints that are intimately tied up with the nature of the nonlinearity in the stochastic term of the equations, and with the choice of the functional spaces in which the problem at hand is well-posed. Our method of proof rests on compactness arguments in an essential way. The second part is devoted to the study of the blowup behavior of solutions to semilinear stochastic partial differential equations with Dirichlet boundary conditions driven by a class of differential operators including (not necessarily symmetric) Lévy processes and diffusion processes, and perturbed by a mixture of Brownian and fractional Brownian motions. Our aim is to understand the influence of the stochastic part and that of the differential operator on the blowup behavior of the solutions. In particular we derive explicit expressions for an upper and a lower bound of the blowup time of the solution and provide a sufficient condition for the existence of global positive solutions. Furthermore, we give estimates of the probability of finite time blowup and for the tail probabilities of an upper bound for the blowup time of the solutions

O principal objetivo deste estudo é descrever modelos para séries temporais de ativos financeiros que sejam robustos às tradicionais hipóteses: distribuição gaussiana e continuidade. O primeiro capítulo está preocupado em apresentar, de uma maneira geral, os conceitos matemáticos mais importantes relacionadas a processos estocásticos e difusões. O segundo capítulo trata de processos de incrementos independentes e estacionários, i.e., processos de Lévy, suas trajetórias estocásticas, propriedades distribucionais e, a relação entre processos markovianos e martingales. Alguns dos resultados apresentados neste capítulo são: a estrutura e as propriedades dos processos compostos de Poisson, medida de Lévy, decomposição de Lévy-Itô e representação de Lévy-Khinchin. O terceiro capítulo mostra como construir processos de Lévy por meio de transformações lineares, inclinação da medida de Lévy e subordina ção. Uma atenção especial é dada aos processos subordinados, tais como os modelos variância gama, normal gaussiana invertida e hiperbólico generalizado. Neste capítulo também é apresentado um exemplo pragmático com dados brasileiros de estimação de parâmetros por meio do método de máxima Verossimilhança. O quarto capítulo é devotado aos modelos multidimensionais e, introduz os conceito de cópula ordinária e de Lévy. Mostra-se que é possível caracterizar a dependência entre os componentes de um processo de Lévy multidimensional por meio da cópula de Lévy. Entre os resultados apresentados estão as generalizações do teorema de Sklar e a família de cópulas de Arquimedes aos processos de Lévy. Este capítulo também apresenta alguns exemplos que utilizam métodos de Monte Carlo, para simular processos de Lévy bidimensionais. / The main objective of this study is to describe models for financial assets time series that are robust to the traditional hypothesis: gaussian distributed and continuity. The first chapter are devoted to introduce the most important mathematical tools related to difusions and stochastic processes in general. The second chapter is concerned in the study of independent and stationary increments, i.e., Lévy processes, their sample paths behavior, distributional properties, and the relation to Markov and martingales processes. Some of the results presented are the structure and properties of a compound Poisson processes, Lévy measure, Lévy-Itô decomposition and Lévy-Khinchin representation. The third chapter demonstrates how to construct Lévy processes via linear transformation, tempering the Lévy measure and subordination. A special attention is given to several types of subordinated processes, comprising the variance gamma, the normal inverse gaussian and the generalized hyperbolic models. A pragmatic example of parameter estimation for brazilian data using the method of maximum likelihood is also given. Chapter four is devoted to multidimensional models, which introduces the notion of ordinary and Lévy copulas. It is shown that modelling via Lévy copula it is possible to characterize the dependence among components of multidimensional Lévy processes. Some of the results presented are generalizations of the Sklar’s theorem and the Archmedian family of copulas for Lévy processes. This chapter also presents some examples using Monte Carlo methods for simulating bidimensional Lévy processes.

Dans le monde économique, les contrats d'options sont très utilisés car ils permettent de se couvrir contre les aléas et les risques dus aux fluctuations des prix des actifs sous-jacents. La détermination du prix de ces contrats est d'une grande importance pour les investisseurs.Dans cette thèse, on s'intéresse aux problèmes d'évaluation des options, en particulier les options Européennes et Quanto sur un actif financier dont le prix est modélisé en multi dimensions par un modèle de diffusion-saut à volatilité stochastique avec sauts (1er cas considère la volatilité sans sauts, dans le 2ème cas les sauts sont pris en compte, finalement dans le 3ème cas, l'actif sous-jacent est sans saut et la volatilité suit un CEV modèle sans saut). Ce modèle permet de mieux prendre en compte certains phénomènes observés dans les marchés. Nous développons des méthodes numériques qui déterminent les valeurs des prix de ces options. On présentera d'abord le modèle qui s'écrit sous la forme d'un système d'équations intégro-différentielles stochastiques "EIDS", et on étudiera l'existence et l'unicité de la solution de ce modèle en fonction de ses coefficients, puis on établira le lien entre le calcul du prix de l'option et la résolution de l'équation Intégro-différentielle partielle (EIDP). Ce lien, qui est basé sur la notion des générateurs infinitésimaux, nous permet d'utiliser différentes méthodes numériques pour l'évaluation des options considérées. Nous introduisons alors l'équation variationnelle associée aux EIDP et démontrons qu'elle admet une unique solution dans un espace de Sobolev avec poids en s'inspirant des travaux de Zhang [106].Nous nous concentrons ensuite sur l'approximation numérique du prix de l'option en considérant le problème dans un domaine borné, et nous utilisons pour la résolution numérique la méthode des éléments finis de type (P1), et un schéma d'Euler-Maruyama, pour se servir, d'une part de la méthode de différences finies en temps, et d'autre part de la méthode de Monté Carlo et la méthode Quasi Monte Carlo. Pour cette dernière méthode nous avons utilisé les suites de Halton afin d'améliorer la vitesse de convergence.Nous présenterons une étude comparative des différents résultats numériques obtenus dans plusieurs cas différents afin d'étudier la performance et l'efficacité des méthodes utilisées. / In the modern economic world, the options contracts are used because they allow to hedge against the vagaries and risks refers to fluctuations in the prices of the underlying assets. The determination of the price of these contracts is of great importance for investors.We are interested in problems of options pricing, actually the European and Quanto options on a financial asset. The price of that asset is modeled by a multi-dimentional jump diffusion with stochastic volatility. Otherwise, the first model considers the volatility as a continuous process and the second model considers it as a jump process. Finally in the 3rd model, the underlying asset is without jump and volatility follows a model CEV without jump. This model allow better to take into account some phenomena observed in the markets. We develop numerical methods that determine the values of prices for these options. We first write the model as an integro-differential stochastic equations system "EIDS", of which we study existence and unicity of solutions. Then we relate the resolution of PIDE to the computation of the option value. This link, which is based on the notion of infinitesimal generators, allows us to use different numerical methods. We therefore introduce the variational equation associated with the PIDE, and drawing on the work of Zhang [106], we show that it admits a unique solution in a weights Sobolev space We focus on the numerical approximation of the price of the option, by treating the problem in a bounded domain. We use the finite elements method of type (P1), and the scheme of Euler-Maruyama, for this serve, on the one hand the finite differences method in time, and on the other hand the method of Monte Carlo and the Quasi Monte Carlo method. For this last method we use of Halton sequences to improve the speed of convergence.We present a comparative study of the different numerical results in many different cases in order to investigate the performance and effectiveness of the used methods.

Economic uncertainty may affect significantly people’s behavior and hence macroeconomic variables. It is thus important to understand how people behave in presence of different kinds of economic risk. The present dissertation focuses therefore on the impact of the uncertainty in capital and labor income on the individual saving behavior. The underlying uncertain variables are here modeled as stochastic processes that each obey a specific stochastic differential equation, where uncertainty stems either from Poisson or Lévy processes. The results on the optimal behavior are derived by maximizing the individual expected lifetime utility. The first chapter is concerned with the necessary mathematical tools, the change-of-variables formula and the Hamilton-Jacobi-Bellman equation under Poisson uncertainty. We extend their possible field of application in order make them appropriate for the analysis of the dynamic stochastic optimization problems occurring in the following chapters and elsewhere. The second chapter considers an optimum-saving problem with labor income, where capital risk stems from asset prices that follow geometric L´evy processes. Chapter 3, finally, studies the optimal saving behavior if agents face not only risk but also uncertain spells of unemployment. To this end, we turn back to Poisson processes, which here are used to model properly the separation and matching process.

本研究顯著地發展時間轉換Lévy過程在最適投資組合的運用性。在連續Lévy過程模型設定下，槓桿效果直接地產生跨期波動度不對稱避險需求，而波動度回饋效果則透過槓桿效果間接地發生影響。另外，關於無窮跳躍Lévy過程模型設定部分，槓桿效果仍扮演重要的影響角色，而波動度回饋效果僅在短期投資決策中發生作用。最後，在本研究所提出之一般化隨機波動度不對稱資產報酬動態模型下，得出在無窮跳躍的資產動態模型設定下，擴散項仍為重要的決定項。 / This study significantly extends the applicability of time-changed Lévy processes to the portfolio optimization. The leverage effect directly induces the intertemporal asymmetric volatility hedging demand, while the volatility feedback effect exerts a minor influence via the leverage effect under the pure-continuous time-changed Lévy process. Furthermore, the leverage effect still plays a major role while the volatility feedback effect just works over the short-term investment horizon under the infinite-jump Lévy process. Based on the proposed general stochastic asymmetric volatility asset return model, we conclude that the diffusion term is an essential determinant of financial modeling for index dynamics given infinite-activity jump structure.

O principal objetivo deste estudo é descrever modelos para séries temporais de ativos financeiros que sejam robustos às tradicionais hipóteses: distribuição gaussiana e continuidade. O primeiro capítulo está preocupado em apresentar, de uma maneira geral, os conceitos matemáticos mais importantes relacionadas a processos estocásticos e difusões. O segundo capítulo trata de processos de incrementos independentes e estacionários, i.e., processos de Lévy, suas trajetórias estocásticas, propriedades distribucionais e, a relação entre processos markovianos e martingales. Alguns dos resultados apresentados neste capítulo são: a estrutura e as propriedades dos processos compostos de Poisson, medida de Lévy, decomposição de Lévy-Itô e representação de Lévy-Khinchin. O terceiro capítulo mostra como construir processos de Lévy por meio de transformações lineares, inclinação da medida de Lévy e subordina ção. Uma atenção especial é dada aos processos subordinados, tais como os modelos variância gama, normal gaussiana invertida e hiperbólico generalizado. Neste capítulo também é apresentado um exemplo pragmático com dados brasileiros de estimação de parâmetros por meio do método de máxima Verossimilhança. O quarto capítulo é devotado aos modelos multidimensionais e, introduz os conceito de cópula ordinária e de Lévy. Mostra-se que é possível caracterizar a dependência entre os componentes de um processo de Lévy multidimensional por meio da cópula de Lévy. Entre os resultados apresentados estão as generalizações do teorema de Sklar e a família de cópulas de Arquimedes aos processos de Lévy. Este capítulo também apresenta alguns exemplos que utilizam métodos de Monte Carlo, para simular processos de Lévy bidimensionais. / The main objective of this study is to describe models for financial assets time series that are robust to the traditional hypothesis: gaussian distributed and continuity. The first chapter are devoted to introduce the most important mathematical tools related to difusions and stochastic processes in general. The second chapter is concerned in the study of independent and stationary increments, i.e., Lévy processes, their sample paths behavior, distributional properties, and the relation to Markov and martingales processes. Some of the results presented are the structure and properties of a compound Poisson processes, Lévy measure, Lévy-Itô decomposition and Lévy-Khinchin representation. The third chapter demonstrates how to construct Lévy processes via linear transformation, tempering the Lévy measure and subordination. A special attention is given to several types of subordinated processes, comprising the variance gamma, the normal inverse gaussian and the generalized hyperbolic models. A pragmatic example of parameter estimation for brazilian data using the method of maximum likelihood is also given. Chapter four is devoted to multidimensional models, which introduces the notion of ordinary and Lévy copulas. It is shown that modelling via Lévy copula it is possible to characterize the dependence among components of multidimensional Lévy processes. Some of the results presented are generalizations of the Sklar’s theorem and the Archmedian family of copulas for Lévy processes. This chapter also presents some examples using Monte Carlo methods for simulating bidimensional Lévy processes.

In this thesis, we present a new existence result for Lévy-type processes. Lévy-type processes behave locally like a Lévy process, but the Lévy triplet may depend on the current position of the process. They can be characterized by their so-called symbol; this is the analogue of the characteristic exponent in the Lévy case. Using a parametrix construction, we prove the existence of Lévy-type processes with a given symbol under weak regularity assumptions on the regularity of the symbol. Applications range from existence results for stable-like processes and mixed processes to uniqueness results for Lévy-driven stochastic differential equations. Moreover, we discuss sufficient conditions for the existence of moments of Lévy-type processes and derive estimates for fractional moments.