In this thesis we investigate numerically SU(2) theories with Dirac—or Majorana—fermions in the adjoint representation. Majorana fermions have historically proven difficult to treat numerically; here, a change of basis is introduced that allows two Majorana fermions to be expressed in terms of one Dirac fermion. This also provides greater insight into the analysis of the properties of theories with Dirac fermions. Attention is focused on the SU(2) theory with a single Dirac flavour (or equivalently two Majorana flavours). Its lattice phase diagram, spectrum, and the anomalous dimension of the chiral condensate are investigated. We observe a long region of constant mass ratios and an anomalous dimension 0.9 ≲ γ∗ ≲ 0.95. The behaviour of the pion mass and the presence of a light scalar in particular point to behaviour that is not traditionally confining; instead the theory appears to lie in or near the conformal window. The topological susceptibility and instanton size distribution are also investigated, for the one-Dirac-flavour theory and additionally the pure-gauge and two-Dirac-flavour (Minimal Walking Technicolor) theories. The properties are found to not depend on number of flavours, indicating a quenching of the fermions in the topology, also consistent with (near-)conformal behaviour (as has previously been reported in studies of other observables for Minimal Walking Technicolor). The code used is described, and a high-performance computing benchmark developed from it is detailed. While the benchmark was originally developed to investigate the performance of different supercomputer architectures for the class of problems we are interested in. Due to the nature of the code on which it is based, it has an unusual flexibility in the demands it may place on machine’s performance characteristics, which may allow it to be applicable to problems outside of lattice physics. The benchmark is used to characterise a number of machines’ relative performance.

In this thesis we study three different modelling frameworks for biological systems of dispersal and combinations thereof. The three frameworks involved are individual-based models, group-level models in the form of partial differential equations (PDEs) and robot swarms. In the first two chapters of the thesis, we present ways of coupling individual based models with PDEs in so-called hybrid models, with the aim of achieving improved performance of simulations. Two classes of such hybrid models are discussed that allow an efficient simulation of multi-species systems of dispersal with reactions, but involve individual resolution for certain species and in certain parts of a computational domain if desired. We generally consider two types of example systems: bacterial chemotaxis and reaction-diffusion systems, and present results in the respective application area as well as general methods. The third chapter of this thesis introduces swarm robotic experiments as an additional tool to study systems of dispersal. In general, those experiments can be used to mimic animal behaviour and to study the impact of local interactions on the group-level dynamics. We concentrate on a target finding problem for groups of robots. We present how PDE descriptions can be adjusted to incorporate the finite turning times observed in the robotic system and that the adjusted models match well with experimental data. In the fourth and last chapter, we consider interactions between robots in the form of hard-sphere collisions and again derive adjusted PDE descriptions. We show that collisions have a significant impact on the speed with which the group spreads across a domain. Throughout these two chapters, we apply a combination of experiments, individual-based simulations and PDE descriptions to improve our understanding of interactions in systems of dispersal.

Cette thèse s'inscrit dans l'étude des systèmes de particules en interaction et plus précisément dans celle des modèles de croissance aléatoire qui représentent un quantité qui grandit au cours du temps et s'étend sur un réseau. Ce type de processus apparaît naturellement quand on regarde la croissance d'un cristal ou bien la propagation d'une épidémie. Cette dernière est bien modélisée par le processus de contact introduit en 1974 par Harris. Le processus de contact est un des plus simples systèmes de particules en interaction présentant une transition de phase et l'on connaît maintenant bien son comportement sur ses phases. De nombreuses questions ouvertes sur ses extensions, notamment celles de formes asymptotiques, ont motivé ce travail. Après la présentation de ce processus et de certaines de ses extensions, nous introduisons et étudions une nouvelle variante: le processus de contact avec vieillissement où les particules ont un âge qui influence leur capacité à donner naissance à leurs voisines. Nous effectuerons pour ce modèle un couplage avec une percolation orientée inspiré de celui de Bezuidenhout-Grimmett et nous montrerons la croissance d'ordre linéaire de ce processus. Dans la dernière partie de la thèse, nous nous intéressons à la preuve d'un théorème de forme asymptotique pour des modèles généraux de croissance aléatoire grâce à des techniques sous-Additives, parfois complexes à mettre en place à cause de la non 'survie presque sûre' de nos modèles. Nous en concluons en particulier que le processus de contact avec vieillissement, le processus de contact en environnement dynamique, la percolation orientée avec immigration hostile, et le processus de contact avec sensibilisation vérifient des résultats de forme asymptotique / This thesis is a contribution to the mathematical study of interacting particles systems which include random growth models representing a spreading shape over time in the cubic lattice. These processes are used to model the crystal growth or the spread of an infection. In particular, Harris introduced in 1974 the contact process to represent such a spread. It is one of the simplest interacting particles systems which exhibits a critical phenomenon and today, its behaviour is well-Known on each phase. Many questions about its extensions remain open and motivated our work, especially the one on the asymptotic shape. After the presentation of the contact process and its extensions, we introduce a new one: the contact process with aging where each particle has an age age that influences its ability to give birth to its neighbours. We build a coupling between our process and a supercritical oriented percolation adapted from Bezuidenhout-Grimmett's construction and we establish the 'at most linear' growth of our process. In the last part of this work, we prove an asymptotic shape theorem for general random growth models thanks to subadditive techniques, which can be complicated in the case of non-Permanent models conditioned to survive. We conclude that the process with aging, the contact process in randomly evolving environment, the oriented percolation with hostile immigration and the bounded modified contact process satisfy asymptotic shape results

Dans cette thèse, on s'intéresse à des systèmes stochastiques modélisant un des phénomènes biologiques les plus mystérieux, les mouvements collectifs de populations. Pour un groupe de N individus, vus comme des particules sans poids ni volume, on étudie deux types de comportements asymptotiques : d'un côté, en temps long, les propriétés d'ergodicité et de flocking, de l'autre, quand le nombre de particules N tend vers l'infini, les phénomènes de propagation du chaos. Le modèle, déterministe, de Cucker-Smale, un modèle cinétique de champ moyen pour une population sans structure hiérarchique, est notre point de départ : les deux premiers chapitres sont consacrés à la compréhension de diverses dynamiques stochastiques qui s'en inspirent, du bruit étant rajouté sous différentes formes. Le troisième chapitre, originellement une tentative d'amélioration de ces résultats, est basé sur la méthode du développement en amas, un outil de physique statistique. On prouve l'ergodicité exponentielle de certains processus non- markoviens à drift non-régulier. Dans la dernière partie, on démontre l'existence d'une solution, unique dans un certain sens, pour un système stochastique de particules associé au modèle chimiotactique de Keller et Segel. / In this thesis, stochastic dynamics modelling collective motions of populations, one of the most mysterious type of biological phenomena, are considered. For a system of N particle-like individuals, two kinds of asymptotic behaviours are studied: ergodicity and flocking properties, in long time, and propagation of chaos, when the number N of agents goes to infinity. Cucker and Smale, deterministic, mean-field kinetic model for a population without a hierarchical structure is the starting point of our journey: the fist two chapters are dedicated to the understanding of various stochastic dynamics it inspires, with random noise added in different ways. The third chapter, an attempt to improve those results, is built upon the cluster expansion method, a technique from statistical mechanics. Exponential ergodicity is obtained for a class of non-Markovian process with non-regular drift. In the final part, the focus shifts onto a stochastic system of interacting particles derived from Keller and Segel 2-D parabolic-elliptic model for chemotaxis. Existence and weak uniqueness are proven.

Dans cette thèse, nous nous intéressons à la modélisation et l'analyse mathématique de certains problèmes liés aux écoulements en suspension.Le premier chapitre concerne la justification du modèle de type transport-Stokes pour la sédimentation de particules sphériques dans un fluide de Stokes où l'inertie des particules est négligée et leur rotation est prise en compte. Ce travail est une extension des résultats antérieurs pour un ensemble plus général de configurations de particules.Le deuxième chapitre concerne la sédimentation d'une distribution d'amas de paires de particules dans un fluide de Stokes. Le modèle dérivé est une équation de transport-Stokes décrivant l'évolution de la position et l'orientation des amas. Nous nous intéressons par la suite au cas où l'orientation des amas est initialement corrélée aux positions. Un résultat d'existence locale et d'unicité pour le modèle dérivé est présenté.Dans le troisième chapitre, nous nous intéressons à la dérivation d'un modèle de type fluide-cinétique pour l'évolution d'un aérosol dans les voies respiratoires. Ce modèle prend en compte la variation du rayon des particules et leur température due à l'échange d'humidité entre l'aérosol et l'air ambiant. Les équations décrivant le mouvement de l'aérosol est une équation de type Vlasov-Navier Stokes couplée avec des équations d'advection diffusion pour l'évolution de la température et la vapeur d'eau dans l'air ambiant.Le dernier chapitre traite de l'analyse mathématique de l'équation de transport-Stokes dérivée au premier chapitre. Nous présentons un résultat d'existence et d'unicité globale pour des densités initiales de type $L^1 cap L^infty$ ayant un moment d'ordre un fini. Nous nous intéressons ensuite à des densités initiales de type fonction caractéristique d'une gouttelette et montrons un résultat d'existence locale et d'unicité d'une paramétrisation régulière de la surface de la gouttelette. Enfin nous présentons des simulations numériques montrant l'aspect instable de la gouttelette. / This thesis is devoted to the modelling and mathematical analysis of some aspects of suspension flows.The first chapter concerns the justification of the transport-Stokes equation describing the sedimentation of spherical rigid particles in a Stokes flow where particles rotation is taken into account and inertia is neglected. This work is an extension of former results for a more general set of particles configurations.The second chapter is dedicated to the sedimentation of clusters of particle pairs in a Stokes flow. The derived model is a transport-Stokes equation describing the time evolution of the position and orientation of the cluster. We also investigate the case where the orientation of the cluster is initially correlated to its position. A local existence and uniqueness result for the limit model is provided.In the third chapter, we propose a coupled fluid-kinetic model taking into accountthe radius growth of aerosol particles due to humidity in the respiratorysystem. We aim to numerically investigate the impact of hygroscopic effects onthe particle behaviour. The air flow is described by the incompressibleNavier-Stokes equations, and the aerosol by a Vlasov-type equation involving the air humidity and temperature, both quantities satisfying a convection-diffusion equation with a source term.The last chapter is dedicated to the analysis of the transport-Stokes equation derived in the first chapter. First we present a global existence and uniqueness result for $L^1cap L^infty$ initial densities with finite first moment. Secondly, we consider the case where the initial data is the characteristic function of a droplet. We present a local existence and uniqueness result for a regular parametrization of the droplet surface. Finally, we provide some numerical computations that show the regularity breakup of the droplet.

Dans cette thèse, nous étudions la généralisation à une infinité de particules d'un modèle hamiltonien décrivant les interactions entre une particule et son environnement. Le milieu est considéré comme une superposition continue de membranes vibrantes. Au bout d'un certain temps, tout se passe comme si la particule était soumise à une force de frottement linéaire. Les équations obtenus pour un grand nombre de particules sont proches des équations de Vlasov. Dans un premier chapitre, on montre d'abord l'existence et l'unicité des solutions puis on s'intéresse à certains régimes asymptotiques; en faisant tendre la vitesse des ondes dans le milieu vers l'infini et en redimensionnant les échelles, on obtient à la limite une équation de Vlasov, on montre que si l'on modifie en plus une fonction paramètrisant le système, on obtient l'équation de Vlasov-Poisson attractive. Dans un deuxième chapitre, on ajoute un terme de diffusion à l'équation. Cela correspond à prendre en compte une agitation brownienne et un frottement linéaire sur les particules. Le principal résultat de ce chapitre est la convergence de la distribution de particules vers une unique distribution stationnaire. On montre la limite de diffusion pour ce nouveau système en faisant tendre simultanément la vitesse de propagation vers l'infini. On obtient une équation plus simple pour la densité spatiale. Dans le chapitre 3, nous montrons la validité des équations déjà étudiées par une limite de champ moyen. Dans le dernier chapitre, on étudie l'asymptotique en temps long de l'équation décrivant l'évolution de la densité spatiale obtenue dans le chapitre 2, des résultats faibles de convergence sont obtenus / The goal of this PhD is to study a generalisation of a model describing the interaction between a single particle and its environment. We consider an infinite number of particles represented by their distribution function. The environment is modelled by a vibrating scalar field which exchanges energy with the particles. In the single particle case, after a large time, the particle behaves as if it were subjected to a linear friction force driven by the environment. The equations that we obtain for a large number of particles are close to the Vlasov equation. In the first chapter, we prove that our new system has a unique solution. We then care about some asymptotic issues; if the wave velocity in the medium goes to infinity, adapting the scaling of the interaction, we connect our system with the Vlasov equation. Changing also continuously a function that parametrizes the model, we also connect our model with the attractive Vlasov-Poisson equation. In the second chapter, we add a diffusive term in our equation. It means that we consider that the particles are subjected to a friction force and a Brownian motion. Our main result states that the distribution function converges to the unique equilibrium distribution of the system. We also establish the diffusive limit making the wave velocity go to infinity at the same time. We find a simpler equation satisfied by the spatial density. In chapter 3, we prove the validity of both equations studied in the two first chapters by a mean field limit. The last chapter is devoted to studying the large time asymptotic properties of the equation that we obtained on the spatial density in chapter 2. We prove some weak convergence results

<div>The rise of artificial intelligence (AI) hinges on the efficient training of modern deep neural networks (DNNs) for non-convex optimization and uncertainty quantification, which boils down to a non-convex Bayesian learning problem. A standard tool to handle the problem is Langevin Monte Carlo, which proposes to approximate the posterior distribution with theoretical guarantees. However, non-convex Bayesian learning in real big data applications can be arbitrarily slow and often fails to capture the uncertainty or informative modes given a limited time. As a result, advanced techniques are still required.</div><div><br></div><div>In this thesis, we start with the replica exchange Langevin Monte Carlo (also known as parallel tempering), which is a Markov jump process that proposes appropriate swaps between exploration and exploitation to achieve accelerations. However, the na\"ive extension of swaps to big data problems leads to a large bias, and the bias-corrected swaps are required. Such a mechanism leads to few effective swaps and insignificant accelerations. To alleviate this issue, we first propose a control variates method to reduce the variance of noisy energy estimators and show a potential to accelerate the exponential convergence. We also present the population-chain replica exchange and propose a generalized deterministic even-odd scheme to track the non-reversibility and obtain an optimal round trip rate. Further approximations are conducted based on stochastic gradient descents, which yield a user-friendly nature for large-scale uncertainty approximation tasks without much tuning costs. </div><div><br></div><div>In the second part of the thesis, we study scalable dynamic importance sampling algorithms based on stochastic approximation. Traditional dynamic importance sampling algorithms have achieved successes in bioinformatics and statistical physics, however, the lack of scalability has greatly limited their extensions to big data applications. To handle this scalability issue, we resolve the vanishing gradient problem and propose two dynamic importance sampling algorithms based on stochastic gradient Langevin dynamics. Theoretically, we establish the stability condition for the underlying ordinary differential equation (ODE) system and guarantee the asymptotic convergence of the latent variable to the desired fixed point. Interestingly, such a result still holds given non-convex energy landscapes. In addition, we also propose a pleasingly parallel version of such algorithms with interacting latent variables. We show that the interacting algorithm can be theoretically more efficient than the single-chain alternative with an equivalent computational budget.</div>