This is a book on numerical methods for singular perturbation problems - in particular stationary convection-dominated convection-diffusion problems. More precisely it is devoted to the construction and analysis of layer-adapted meshes underlying these numerical methods. An early important contribution towards the optimization of numerical methods by means of special meshes was made by N.S. Bakhvalov in 1969. His paper spawned a lively discussion in the literature with a number of further meshes being proposed and applied to various singular perturbation problems. However, in the mid 1980s this development stalled, but was enlivend again by G.I. Shishkin's proposal of piecewise- equidistant meshes in the early 1990s. Because of their very simple structure they are often much easier to analyse than other meshes, although they give numerical approximations that are inferior to solutions on competing meshes. Shishkin meshes for numerous problems and numerical methods have been studied since and they are still very much in vogue. With this contribution we try to counter this development and lay the emphasis on more general meshes that - apart from performing better than piecewise-uniform meshes - provide a much deeper insight in the course of their analysis. In this monograph a classification and a survey are given of layer-adapted meshes for convection-diffusion problems. It tries to give a comprehensive review of state-of-the art techniques used in the convergence analysis for various numerical methods: finite differences, finite elements and finite volumes. While for finite difference schemes applied to one-dimensional problems a rather complete convergence theory for arbitrary meshes is developed, the theory is more fragmentary for other methods and problems and still requires the restriction to certain classes of meshes.

Diese Arbeit behandelt die Asymptotik von Austritts- und Übergangszeiten für gewisse schwach zeitinhomogene Diffusionsprozesse. Darauf basierend wird ein probabilistischer Begriff der stochastischen Resonanz (SR) studiert. Techniken der großen Abweichungen spielen eine zentrale Rolle. Im ersten Teil der Arbeit (Kapitel 1-3) werden Resultate aus der Theorie der großen Abweichungen für zeithomogene Diffusionen rekapituliert. Es werden die klassischen Resultate von Freidlin und Wentzell und Erweiterungen dieser Theorie präsentiert, und es wird an das Kramers''sche Austrittszeitengesetz erinnert. Teil II befasst sich mit dem Phänomen der SR, d.h. mit Periodizitätseigenschaften von Diffusionen. In Kapitel 4 werden physikalische Maße zur Messung der Periodizität diskutiert. Deren Nachteile legen es nahe, einem alternativen, probabilistischen Ansatz zu folgen, der hier behandelt wird. Das 5. Kapitel dient der Herleitung eines gleichmäßigen Prinzips der großen Abweichungen für Diffusionen mit schwach zeitabhängigem, periodischem Drift. Die Gleichmäßigkeit des Prinzips ermöglicht die exakte Bestimmung exponentieller Übergangsraten in Kapitel 6, das die zentralen Ergebnisse des 2. Teils beinhaltet. Hierdurch wird die Maximierung gewisser Übergangswahrscheinlichkeiten ermöglicht, was zum in Kapitel 7 studierten Resonanzbegriff führt. Teil III der Arbeit setzt sich mit der Asymptotik von Austrittszeiten sogenannter selbststabilisierender Diffusionen auseinander. In Kapitel 8 wird der Zusammenhang zwischen interagierenden Teilchensystemen und selbststabilisierenden Diffusionen erläutert und die Existenz- und Eindeutigkeitsfrage behandelt. Das 9. Kapitel dient dem Studium der großen Abweichungen dieser Klasse von Diffusionen. In Kapitel 10 wird das Kramers''sche Austrittszeitengesetz auf selbststabilisierende Diffusionen übertragen, und in Kapitel 11 wird der Einfluß der selbststabilisierenden Komponente auf das Austrittszeitengesetz illustriert. / In this thesis, we study the asymptotic behavior of exit and transition times of certain weakly time inhomogeneous diffusion processes. Based on these asymptotics, a probabilistic notion of stochastic resonance (SR) is investigated. Large deviations techniques play the key role throughout this work. In the first part (Chapters 1-3) we recall the large deviations theory for time homogeneous diffusions. We present the classical results due to Freidlin and Wentzell and extensions thereof, and we remind of Kramers'' exit time law. Part II deals with the phenomenon of stochastic resonance. That is, we study periodicity properties of diffusion processes. In Chapter 4 we explain the paradigm of stochastic resonance and discuss physical notions of measuring periodicity of diffusions. Their drawbacks suggest to follow an alternative probabilistic approach, which is treated in this work. In Chapter 5 we derive a large deviations principle for diffusions subject to a weakly time dependent periodic drift term. The uniformity of the obtained large deviations bounds w.r.t. the system''s parameters plays a key role for the treatment of transition time asymptotics in Chapter 6, which contains the main result of the second part. The exact exponential transition rates obtained here allow for maximizing transition probabilities, which finally leads to the announced probabilistic notion of resonance studied in Chapter 7. In the third part we investigate the exit time asymptotics of a certain class of so-called self-stabilizing diffusions. In Chapter 8 we explain the connection between interacting particle systems and self-stabilizing diffusions, and we address the question of existence. The following Chapter 9 is devoted to the study of the large deviations behavior of these diffusions. In Chapter 10 Kramers'' exit law is carried over to our class of self-stabilizing diffusions. Finally, the influence of self-stabilization is illustrated in Chapter 11.

Das globale Klimasystem ist ein ausgesprochen komplexes und hochgradig nichtlineares System mit einer Vielzahl von Einflüssen und Interaktionen zwischen Variablen und Parametern. Komplementär zu der Beschreibung des Systems mit globalen Klimamodellen, kann Klima anhand der Interaktionsstruktur des Gesamtsystems durch Netzwerke beschrieben werden. Statt Details so genau wie möglich zu modellieren, werden hier Zeitreihendaten verwendet um zugrundeliegende Strukturen zu finden. Die Herausforderung liegt dann in der Interpretation dieser Strukturen. Um mich der Frage nach der Interpretation von Netzwerkmaßen zu nähern, suche ich nach einem allgemeinen Zusammenhang zwischen Eigenschaften der Netzwerktopologie und Eigenschaften des zugrundeliegenden physikalischen Systems. Dafür werden im Wesentlichen zwei Methoden entwickelt, die auf der Analyse von Temperaturentwicklungen gemäß der Advektions-Diffusions-Gleichung (ADE) basieren. Für die erste Methode wird die ADE mit offenen Randbedingungen und δ-peak Anfangsbedingungen gelöst. Die resultierenden lokalen Temperaturprofile werden verwendet um eine Korrelationsfunktion und damit ein Netzwerk zu definieren. Diese Netzwerke werden analysiert und mit Klimanetzen aus Daten verglichen. Die zweite Methode basiert auf der Diskretisierung der stochastischen ADE. Die resultierende lineare, stochastische Rekursionsgleichung wird verwendet um eine Korrelationsmatrix zu definieren, die nur von der Übergangsmatrix und der Varianz des stochastischen Störungsterms abhängt. Ich konstruiere gewichtete und ungewichtete Netzwerke für vier verschiedene Fälle und schlage Netzwerkmaße vor, die zwischen diesen Systemen zu unterscheiden helfen, wenn nur das Netzwerk und die Knotenpositionen gegeben sind. Die präsentierten Rekonstruktionsmethoden generieren Netzwerke, die konzeptionell und strukturell Klimanetzwerken ähneln und können somit als "proof of concept" der Methode der Klimanetzwerke, sowie als Interpretationshilfe betrachtet werden. / The earth’s climate is an extraordinarily complex, highly non-linear system with a multitude of influences and interactions between a very large number of variables and parameters. Complementary to the description of the system using global climate models, in recent years, a description based on the system’s interaction structure has been developed. Rather than modelling the system in as much detail as possible, here time series data is used to identify underlying large scale structures. The challenge then lies in the interpretation of these structures. In this thesis I approach the question of the interpretation of network measures from a general perspective, in order to derive a correspondence between properties of the network topology and properties of the underlying physical system. To this end I develop two methods of network construction from a velocity field, using the advection-diffusion-equation (ADE) for temperature-dissipation in the system. For the first method, the ADE is solved for δ-peak-shaped initial and open boundary conditions. The resulting local temperature profiles are used to define a correlation function and thereby a network. Those networks are analysed and compared to climate networks from data. Despite the simplicity of the model, it captures some of the most salient features of climate networks. The second network construction method relies on a discretisation of the ADE with a stochastic term. I construct weighted and unweighted networks for four different cases and suggest network measures, that can be used to distinguish between the different systems, based on the topology of the network and the node locations. The reconstruction methods presented in this thesis successfully model many features, found in climate networks from well-understood physical mechanisms. This can be regarded as a justification of the use of climate networks, as well as a tool for their interpretation.

This thesis deals with a newly-developed model for the spatiotemporal calcium dynamics within presynaptic terminals. The model is based on single-protein kinetics and has been used to successfully describe different neuron types such as pyramidal neurons in the rat neocortex and the Calyx of Held of neurons from the rat brainstem. A limited number of parameters had to be adjusted to fluorescence measurements of the calcium concentration. These values can be interpreted as a prediction of the model, and in particular the protein densities can be compared to independent experiments. The contribution of single proteins to the total calcium dynamics has been analysed in detail for voltage-dependent calcium channel, plasma-membrane calcium ATPase, sodium-calcium exchanger, and endogenous as well as exogenous buffer proteins. The model can be used to reconstruct the unperturbed calcium dynamics from measurements using fluorescence indicators. The calcium response to different stimuli has been investigated in view of its relevance for synaptic plasticity. This work provides a first step towards a description of the complete synaptic transmission using single-protein data.

Dans cette thèse on étudie des schémas numériques pour des processus<br />/X/ à coefficients discontinus. Un premier schéma pour le cas<br />unidimensionnel utilise les Équations Différentielles Stochastiques<br />avec Temps Local. En effet en dimension un les processus /X/ sont<br />solutions de telles équations. On construit une grille sur la droite<br />réelle, qu'une bijection adéquate transforme en une grille uniforme<br />de pas /h/. Cette bijection permet de transformer /X/ en /Y/ qui se<br />comporte localement comme un Skew Brownian Motion, pour lequel on<br />connaît les probabilités de transition sur une grille uniforme, et le<br />temps moyen passé sur chaque cellule de cette grille. Une marche<br />aléatoire peut alors être construite, qui converge vers /X/ en racine<br />de /h/. Toujours dans le cas unidimensionnel on propose un deuxième<br />schéma plus général. On se donne une grille non uniforme sur la<br />droite réelle, dont les cellules ont une taille proportionnelle à<br />/h/. On montre qu'on peut relier les probabilités de transition de<br />/X/ sur cette grille, ainsi que le temps moyen passé par /X/ sur<br />chacune de ses cellules, à des solutions de problèmes d'EDP<br />elliptiques ad hoc. Une marche aléatoire en temps et en espace est<br />ainsi construite, qui permet d'approcher /X/ à nouveau en racine de<br />/h/. Ensuite on présente des pistes pour adapter cette dernière<br />approche au cas bidimensionnel et les problèmes que cela soulève.<br />Enfin on illustre par des exemples numériques les schémas étudiés.

No description available.

Cette thèse est consacrée à l’étude des systèmes d’équations différentielles ordinaires, et ceux aux dérivées partielles paraboliques issus de modèles de dynamique des populations et de la biologie. L’objectif principal est de faire l’analyse mathématique, la simulation numérique ainsi que l’identification des diffusions croisées dans les modèles construits. Nous présentons d’abord un système de réaction-diffusion modélisant la croissance de plantes en compétition spatiale dans un milieu saturé. Nous effectuons par la suite l’étude théorique et numérique de tels systèmes, ainsi que l’étude des problèmes d’identification des termes de diffusions croisées. Ensuite, nous proposons un modèle proie-prédateur de type Leslie-Gower modifié avec une fonction de réponse de type Crowley-Martin. Nous étudions dans un premier temps la dynamique temporelle globale du modèle considéré, et nous présentons des simulations numériques pour illustrer les résultats théoriques. En outre, nous introduisons la dimension spatiale dans le modèle dynamique considéré, et nous effectuons une analyse théorique complète de la dynamique spatio-temporelle du modèle. / This thesis is devoted to the study of ordinary differential systems, and systems of non linear parabolic PDEs resulting from models of population dynamics and biology. The main objective is to perform mathematical analysis, numerical simulations, and identification of cross-diffusion in built models. We first present a reaction-diffusion system that models the spatial competition of plants in a saturated environment. We then perform a theoretical and a numerical study of such systems, and handle the identification of cross-diffusion problem. Secondly, we propose a modified Leslie-Gower-type predator-prey model with a Crowley-Martin type functional response. Within this context, we study the global temporal dynamics of the considered model, and present numerical simulations as illustration of the theoretical results. Finally, we introduce the spatial dimension in the previous dynamical model, and perform a comprehensive theoretical analysis of the spatio-temporal model.

This thesis deals with a newly-developed model for the spatiotemporal calcium dynamics within presynaptic terminals. The model is based on single-protein kinetics and has been used to successfully describe different neuron types such as pyramidal neurons in the rat neocortex and the Calyx of Held of neurons from the rat brainstem. A limited number of parameters had to be adjusted to fluorescence measurements of the calcium concentration. These values can be interpreted as a prediction of the model, and in particular the protein densities can be compared to independent experiments. The contribution of single proteins to the total calcium dynamics has been analysed in detail for voltage-dependent calcium channel, plasma-membrane calcium ATPase, sodium-calcium exchanger, and endogenous as well as exogenous buffer proteins. The model can be used to reconstruct the unperturbed calcium dynamics from measurements using fluorescence indicators. The calcium response to different stimuli has been investigated in view of its relevance for synaptic plasticity. This work provides a first step towards a description of the complete synaptic transmission using single-protein data.

Reaction-diffusion models have been widely successful in explaining a large variety of patterning phenomena in biology ranging from embryonic development to cancer growth and angiogenesis. Firstly proposed by Alan Turing in 1952 and applied to a simple two-component system, reaction-diffusion models describe spontaneous spatial pattern formation, driven purely by interactions of the system components and their diffusion in space. Today, access to unprecedented amounts of quantitative biological data allows us to build and test biochemically accurate reaction-diffusion models of intracellular processes. However, any increase in model complexity increases the number of unknown parameters and thus the computational cost of model analysis. To efficiently characterize the behavior and robustness of models with many unknown parameters is, therefore, a key challenge in systems biology. Here, we propose a novel computational framework for efficient high-dimensional parameter space characterization of reaction-diffusion models. The method leverages the $L_p$-Adaptation algorithm, an adaptive-proposal statistical method for approximate high-dimensional design centering and robustness estimation. Our approach is based on an oracle function, which describes for each point in parameter space whether the corresponding model fulfills given specifications. We propose specific oracles to estimate four parameter-space characteristics: bistability, instability, capability of spontaneous pattern formation, and capability of pattern maintenance. We benchmark the method and demonstrate that it allows exploring the ability of a model to undergo pattern-forming instabilities and to quantify model robustness for model selection in polynomial time with dimensionality. We present an application of the framework to reconstituted membrane domains bearing the small GTPase Rab5 and propose molecular mechanisms that potentially drive pattern formation.

<p>Graph diffusion is the process of spreading information from one or few nodes to the rest of the graph through edges. The resulting distribution of the information often implies latent structure of the graph where nodes more densely connected can receive more signal. This makes graph diffusions a powerful tool for local clustering, which is the problem of finding a cluster or community of nodes around a given set of seeds. Most existing literatures on using graph diffusions for local graph clustering are linear diffusions as their dynamics can be fully interpreted through linear systems. They are also referred as eigenvector, spectral, or random walk based methods. While efficient, they often have difficulty capturing the correct boundary of a target label or target cluster. On the contrast, maxflow-mincut based methods that can be thought as 1-norm nonlinear variants of the linear diffusions seek to "improve'' or "refine'' a given cluster and can often capture the boundary correctly. However, there is a lack of literature to adopt them for problems such as community detection, local graph clustering, semi-supervised learning, etc. due to the complexity of their formulation. We addressed these issues by performing extensive numerical experiments to demonstrate the performance of flow-based methods in graphs from various sources. We also developed an efficient LocalGraphClustering Python Package that allows others to easily use these methods in their own problems. While studying these flow-based methods, we find that they cannot grow from small seed set. Although there are hybrid procedures that incorporate ideas from both linear diffusions and flow-based methods, they have many hard to set parameters. To tackle these issues, we propose a simple generalization of the objective function behind linear diffusion and flow-based methods which we call generalized local graph min-cut problem. We further show that by involving p-norm in this cut problem, we can develop a nonlinear diffusion procedure that can find local clusters from small seed set and capture the correct boundary simultaneously. Our method can be thought as a nonlinear generalization of the Anderson-Chung-Lang push procedure to approximate a personalized PageRank vector efficiently and is a strongly local algorithm-one whose runtime depends on the size of the output rather than the size of the graph. We also show that the p-norm cut functions improve on the standard Cheeger inequalities for linear diffusion methods. We further extend our generalized local graph min-cut problem and the corresponding diffusion solver to hypergraph-based machine learning problems. Although many methods for local graph clustering exist, there are relatively few for localized clustering in hypergraphs. Moreover, those that exist often lack flexibility to model a general class of hypergraph cut functions or cannot scale to large problems. Our new hypergraph diffusion method on the other hand enables us to compute with a wide variety of cardinality-based hypergraph cut functions and still maintains the strongly local property. We also show that the clusters found by solving the new objective function satisfy a Cheeger-like quality guarantee.</p> <p>Besides clustering, recent work on graph-based learning often focuses on node embeddings and graph neural networks. Although these GNN based methods can beat traditional ones especially when node attributes data is available, it is challenging to understand them because they are highly over-parameterized. To solve this issue, we propose a novel framework that combines topological data analysis and diffusion to transform the complex prediction space into human understandable pictures. The method can be applied to other datasets not in graph formats and scales up to large datasets across different domains and enable us to find many useful insights about the data and the model.</p>