This paper describes the analysis tool DSPNexpress which has been developed at the Technische Universität Berlin since 1991. The development of DSPNexpress has been motivated by the lack of a powerful software package for the numerical solution of deterministic and stochastic Petri nets (DSPNs) and the complexity requirements imposed by evaluating memory consistency models for multicomputer systems. The development of DSPNexpress has gained by the author's experience with the version 1.4 of the software package GreatSPN. However, opposed to GreatSPN, the software architecture of DSPNexpress is particularly tailored to the numerical evaluation of DSPNs. Furthermore, DSPNexpress contains a graphical interface running under the X11 window system. To the best of the author's knowledge, DSPNexpress is the first software package which contains an efficient numerical algorithm for computing steady-state solutions of DSPNs.

For both randomized clinical trials and prospective cohort studies, the Cox regression model is a powerful tool for evaluating the effect of a treatment or an explanatory variable on time-to-event outcome. This method assumes proportional hazards over time. Systematic approaches to efficiently evaluate non-proportionality and to model data in the presence of non-proportionality are investigated. Six graphical methods are assessed to verify the proportional hazards assumption based on characteristics of the survival function, cumulative hazard, or the feature of residuals. Their performances are empirically evaluated with simulations by checking their ability to be consistent and sensitive in detecting proportionality or non-proportionality. Two-sample data are generated in three scenarios of proportional hazards and five types of alternatives (that is, non-proportionality). The usefulness of these graphical assessment methods depends on the event rate and type of non-proportionality. Three numerical (statistical testing) methods are compared via simulation studies to investigate the proportional hazards assumption. In evaluating data for proportionality versus non-proportionality, the goal is to test a non-zero slope in a regression of the variable or its residuals on a specific function of time, or a Kolmogorov-type supremum test. Our simulation results show that statistical test performance is affected by the number of events, event rate, and degree of divergence of non-proportionality for a given hazards scenario. Determining which test will be used in practice depends on the specific situation under investigation. Both graphical and numerical approaches have benefits and costs, but they are complementary to each other. Several approaches to model and summarize non-proportionality data are presented, including non-parametric measurements and testing, semi-parametric models, and a parametric approach. Some illustrative examples using simulated data and real data are also presented. In summary, we present a systemic approach using both graphical and numerical methods to identify non-proportionality, and to provide numerous modeling strategies when proportionality is violated in time-to-event data.

<p>High-dimensional regression problems are ubiquitous in science and engineering. Deep learning has been a critical tool for solving a wide range of high-dimensional problems with surprising performance. Even though in theory neural networks have good properties in terms of approximation and optimization, numerically obtaining an accurate neural network solution is a challenging problem due to the highly non-convex objective function and implicit bias of least square optimization. In this dissertation, we mainly discuss two topics involving the high dimensional regression using efficient deep learning algorithms. These two topics include solving PDEs with high dimensional domains and data-driven dynamical modeling. </p> <p><br></p> <p>In the first topic, we aim to develop an efficient solver for PDE problems. Firstly, we focus on neural network structures to increase efficiency. We propose a data-driven activation function called reproducing activation function which can reproduce traditional approximation tools and enable faster convergence of deep neural network training with smaller parameter cost. Secondly, we target the application of neural networks to mitigate the numerical issues that hamper the traditional approach. As an example, we develop a neural network solver for elliptic PDEs on unknown manifolds and verify its effectiveness for the large-scale problem. </p> <p><br></p> <p>In the second topic, we aim to enhance the accuracy of learning the dynamical system from data by incorporating the prior. In the missing dynamics problem, taking advantage of known partial dynamics, we propose a framework that approximates a map that takes the memories of the resolved and identifiable unresolved variables to the missing components in the resolved dynamics. With this framework, we achieve a low error to predict the missing component, enabling the accurate prediction of the resolved variables. In the recovering Hamiltonian dynamics, by the energy conservation property, we learn the conserved Hamiltonian function instead of its associated vector field. To better learn the Hamiltonian from the stiff dynamics, we identify and splits the</p> <p>training data into stiff and nonstiff portions, and adopt different learning strategies based on the classification to reduce the training error. </p>

>Magister Scientiae - MSc / We present the Black-Scholes Merton partial differential equation (BSMPDE) and its analytical solution. We present the Black-Scholes option pricing model and list some limitations of this model. We also present a nonlinear model (the Frey-Patie model) that may improve on one of these limitations. We apply various numerical methods on the BSMPDE and run simulations to compare which method performs best in approximating the value of a European put option based on the maximum errors each method produces when we vary some parameters like the interest rate and the volatility. We re-apply the same finite difference methods on the nonlinear model. / 2025

La résolution numérique directe des problèmes de transport de photons en interaction avec un milieu matériel est très coûteuse en mémoire et temps CPU. Pour pallier ce problème, une méthode consiste à construire des modèles réduits dont la résolution est moins coûteuse. La littérature abonde de ce genre de modèles : modèles probabilistes (Monte-Carlo), modèles aux moments (M₁, PN), modèles aux ordonnées discrètes (SN), modèles de diffusion... Dans cette thèse, nous nous intéressons aux modèles PN dans lesquels l'opérateur de transport est approché par projections sur une base tronquée d'harmoniques sphériques. Ces modèles ont l'avantage d'être arbitrairement précis sur la dimension angulaire et ne présentent pas les défauts connus des autres méthodes (bruit stochastique, "effets de raies") pouvant briser les éventuelles symétries du problème. Ce dernier point est capital pour la simulation d'expériences de fusion par confinement inertiel (FCI) où la symétrie sphérique joue un rôle important dans la précision des résultats. Nous étudions donc dans cette thèse la structure mathématique des modèles PN ainsi que leur discrétisation dans le cas d'une géométrie 1D sphérique.Nous commençons par le cas du transport linéaire dans le vide. Même dans ce cas simple, les équations du modèle PN contiennent des termes sources d'origine géométrique dont la discrétisation s'avère délicate. Jusqu'à présent, les différents schémas utilisés étaient insatisfaisants pour les raisons suivantes : (1) mauvais comportement au voisinage de r = 0 (phénomène de "flux-dip"), (2) non préservation des équilibres stationnaires, (3) pas de preuve formelle de stabilité. À la lumière de récents travaux, nous proposons une nouvelle discrétisation qui capture exactement les états d'équilibres. Nous démontrons en particulier la stabilité en norme L² du schéma. Nous étendons par la suite ce schéma au cas du transport de photons dans un milieu matériel figé et nous nous intéressons au comportement du schéma en limite diffusion (propriété "asymptotic-preserving").Dans un second temps, nous nous intéressons au couplage entre rayonnement et hydrodynamique. Devant l'absence de consensus sur les modèles "transport" d'hydrodynamique radiative issus de la littérature, nous établissons une étude comparative de ceux-ci basée sur leurs propriétés mathématiques. Nous nous intéressons particulièrement aux propriétés suivantes : (1) conservation de l'énergie et de l'impulsion, (2) précision des effets comobiles, (3) existence d'une entropie mathématiques compatible et (4) restitution de la limite diffusion. Notre étude se réduit aux modèles dits "mixed-frame" et une attention particulière est toujours portée sur l'approximation "PN" de l'opérateur de transport. Nous identifions des défauts (conservation ou entropie) sur des modèles existants et proposons une correction entropique conduisant à un modèle PN satisfaisant toutes les propriétés mathématiques listées ci-dessus. / Computational costs for direct numerical simulations of photon transport problemsare very high in terms of CPU time and memory. One way to tackle this issue is todevelop reduced models that a cheaper to solve numerically. There exists number of these models : moments models, discrete ordinates models (SN), diffusion-like models... In this thesis, we focus on PN models in which the transport operator is approached by mean of a truncated development on the spherical harmonics basis. These models are arbitrary accurate in the angular dimension and are rotationnaly invariants (in multiple space dimensions). The latter point is fundamental when one wants to simulate inertial confinment fusion (ICF) experiments where the spherical symmetry plays an important part in the accuracy of the numerical solutions. We study the mathematical structure of the PN models and construct a new numerical method in the special case of a one dimensionnal space dimension with spherical symmetry photon transport problems. We first focus on a linear transport problem in the vacuum. Even in this simple case, it appears in the PN equations geometrical source terms that are stiff in the neighborhood of r = 0 and thus hard to discretise. Existing numerical methods are not satisfactory for multiple reasons : (1) unaccuracy in the neighborhood of r = 0 ("flux-dip"), (2) do not capture steady states (well-balanced scheme), (3) no stability proof. Following recent works, we develop a new well-balanced scheme for which we show the L² stability. We then extend the scheme for photon transport problems within a no moving media, the linear Boltzmann equation, and interest ourselves on its behavior in the diffusion limit (asymptotic-preserving property). In a second part, we consider radiation hydrodynamics problems. Since modelisation of these problems is still under discussion in the litterature, we compare a set of existing models by mean of mathematical analysis and establish a hierarchy. For each model, we focus on the following mathematical properties : (1) energy and impulsion conservation, (2) accuracy of the comobile effects, (3) existence of a mathematical entropy and (4) behavior in the diffusion limit. Our study reduces to « laboratory frame » models and we are still interested in the PN approximation of the transport operator. We identify defects in entropy structure of existing models and propose an entroy correction which leads to PN-based radiation hydrodynamics models which satisfy all the properties listed above.

No description available.

Low frequency acoustic room behaviour can be approximated using numerical methods. Traditionally, music studio control rooms are built with complex geometries, making their eigenmodes difficult to predict mathematically. Hence, a summation-by-parts method with simultaneous-approximation-terms is derived to approximate the time dependent acoustic wave equation in three dimensions. The derived model is limited to rectangular prismatic rooms but planned to be expanded to handle complex geometries in the future. Semi-reflecting boundary conditions are used, corresponding to tabulated reflection and absorption properties of real. walls. Two speakers are modeled as omnidirectional point sources placed on a boundary, to mimic common studio setups. Through tests and examination of eigenvalues of the matrices in the method, conditions for stability and reflection coefficients are derived. Simulations of sound pressure distribution produced by the model correlate well to room mode theory, suggesting the model to be accurate in the application of predicting low frequency acoustic room behaviour. However, the convergence rate of the model turns out to be lower than expected when point sources are introduced. Future steps towards applying the model to real music studio control rooms include modeling the walls as changes in density and wave speed rather than boundaries of the domain. This would potentially allow more complex geometries to be modeled within a larger, rectangular domain.

Calcium signaling has been identified with triggering of gene transcriptions associated with learning and neuroprotection in neurons. Studies indicate that dysregulation of calcium signaling is correlated with severe Alzheimer Disease pathologies. A stable calcium wave or signal arising from triggers in dendritic synapses needs to reach soma with constant amplitude for proper functioning of neurons. In this study, we introduce "CalciumSim", a calcium dynamics simulator which works on dimensionally reduced model. Numerical analysis is conducted to obtain the best configuration of neuron geometry to make the code efficient and fast. Alongside, biologically important insights are derived by modulating and changing parameters of the simulation. The ability of "CalciumSim" to work with real neuron geometries allows user to study calcium signalling in a realistic model. / Mathematics

No description available.

The accuracy of moment method solutions to electromagnetic scattering problems has been studied by many researchers. Error bounds for the moment method have been obtained in terms of Sobolev norms of the current solution. Motivated by the historical origins of Sobolev spaces as energy spaces, it is shown that the Sobolev norm used in these bounds is equivalent to the forward scattering amplitude, for the case of 2D scattering from a PEC circular cylinder. A slightly weaker relationship is obtained for 3D scattering from a PEC sphere. These results provide a physical meaning for abstract solution error bounds in terms of the power radiated by the error in the current solution. It is further shown that bounds on the Sobolev norm of the current error imply a bound on the error in the computed backscattering amplitude. Since Sobolev-based error bounds do not provide the actual error in a solution nor identify its source, the error in typical moment method scattering solutions for smooth cylindrical geometries is analyzed. To quantify the impact of mesh element size, approximate integration of moment matrix elements, and geometrical discretization error on the accuracy of computed surface currents and scattering amplitudes, error estimates are derived analytically for the circular cylinder. These results for the circular cylinder are empirically compared to computed error values for other smooth scatterer geometries, with consistent results obtained. It is observed that moment method solutions to the magnetic field integral equation are often less accurate for a given grid than corresponding solutions to the electric field integral equation. Building from the error analysis, the cause of this observation is proposed to be the identity operator in the magnetic formulation. A regularization of the identity operator is then derived that increases the convergence rate of the discretized 2D magnetic field integral equation by three orders.