Searcharxiv⌕ Search

arXiv subjects

Tobias Schaefer

Publications and source records attributed to Tobias Schaefer.

14 recordsLinked to original sources
arXiv2025

Probabilistic Rule Models as Diagnostic Layers: Interpreting Structural Concept Drift in Post-Crisis Finance

Machine learning models used for high-stakes predictions in domains like credit risk face critical degradation due to concept drift, requiring robust and transparent adaptation mechanisms. We propose an architecture, where a dedicated correction layer is employed to efficiently capture systematic shifts in predictive scores when a model becomes outdated. The key element of this architecture is the design of a correction layer using Probabilistic Rule Models (PRMs) based on Markov Logic Networks, which guarantees intrinsic interpretability through symbolic, auditable rules. This structure transforms the correction layer from a simple scoring mechanism into a powerful diagnostic tool capable of isolating and explaining the fundamental changes in borrower riskiness. We illustrate this diagnostic capability using Fannie Mae mortgage data, demonstrating how the interpretable rules extracted by the correction layer successfully explain the structural impact of the 2008 financial crisis on specific population segments, providing essential insights for portfolio risk management and regulatory compliance.

q-fin.RM↗
arXiv2024

Instantons, fluctuations and singularities in the supercritical stochastic nonlinear Schroedinger equation

Recently, Josserand et al. proposed a stochastic nonlinear Schroedinger model for finite-time singularity-mediated turbulence [Phys. Rev. Fluids 5, 054607 (2020)]. Here, we use instanton calculus to quantify the effect of extreme fluctuations on the statistics of the energy dissipation rate. While the contribution of the instanton alone is insufficient, we obtain excellent agreement with direct simulations when including Gaussian fluctuations and the corresponding zero mode. Fluctuations are crucial to obtain the correct scaling when quasi-singular events govern the turbulence statistics.

physics.flu-dyn↗
arXiv2021

GNOME and LBM Model Evaluation on Ocean Oil Spill Far-Field Impacts to Highly Sensitive Areas

In case of an ocean oil spill, there are certain areas, e.g. shrimp farms, which are highly sensitive to small amounts of oil pollution while they are geographically far from the spill. We investigate the Lattice Boltzmann Method (LBM) and GNOME, a tool developed and used by NOAA, in terms of far-field impacts to sensitive areas. We present our simulation results of both models in limited scale (a sub area of Gulf of Mexico) under the same oil spill condition using real ocean current data from the Unified Wave Interface-Coupled Model (UWIN-CM). Our study shows that the kinetic theory based LBM model outperforms the stochastic particle based GNOME model in accuracy and computation time due to their fundamental difference in representation of oil pollution advection and diffusion mechanisms. We propose LBM as a viable alternative to the Lagrangian particle calculation component of GNOME in modeling the far-field impacts to highly sensitive areas.

physics.ao-ph↗
arXiv2017

Assessing density functionals using many body theory for hybrid perovskites

Which density functional is the "best" for structure simulations of a particular material? A concise, first principles, approach to answer this question is presented. The random phase approximation (RPA)--- an accurate many body theory--- is used to evaluate various density functionals. To demonstrate and verify the method, we apply it to the hybrid perovskite MAPbI$_3$, a promising new solar cell material. The evaluation is done by first creating finite temperature ensembles for small supercells using RPA molecular dynamics, and then evaluating the variance between the RPA and various approximate density functionals for these ensembles. We find that, contrary to recent suggestions, van der Waals functionals do not improve the description of the material, whereas hybrid functionals and the SCAN (strongly constrained appropriately normed) density functional yield very good agreement with the RPA. Finally, our study shows that in the room temperature tetragonal phase of MAPbI$_3$, the molecules are preferentially parallel to the shorter lattice vectors but reorientation on ps timescales is still possible.

cond-mat.mtrl-sci↗
arXiv2016

Long Term Effects of Small Random Perturbations on Dynamical Systems: Theoretical and Computational Tools

Small random perturbations may have a dramatic impact on the long time evolution of dynamical systems, and large deviation theory is often the right theoretical framework to understand these effects. At the core of the theory lies the minimization of an action functional, which in many cases of interest has to be computed by numerical means. Here we review the theoretical and computational aspects behind these calculations, and propose an algorithm that simplifies the geometric minimum action method to minimize the action in the space of arc-length parametrized curves. We then illustrate this algorithm's capabilities by applying it to various examples from material sciences, fluid dynamics, atmosphere/ocean sciences, and reaction kinetics. In terms of models, these examples involve stochastic (ordinary or partial) differential equations with multiplicative or degenerate noise, Markov jump processes, and systems with fast and slow degrees of freedom, which all violate detailed balance, so that simpler computational methods are not applicable.

math.NA↗
arXiv2015

A state vector algebra for algorithmic implementation of second-order logic

We present a mathematical framework for mapping second-order logic relations onto a simple state vector algebra. Using this algebra, basic theorems of set theory can be proven in an algorithmic way, hence by an expert system. We illustrate the use of the algebra with simple examples and show that, in principle, all theorems of basic set theory can be recovered in an elementary way. The developed technique can be used for an automated theorem proving in the 1st and 2nd order logic.

cs.AI↗
arXiv2013

Averaging and spectral properties for the 2D advection-diffusion equation in the semi-classical limit for vanishing diffusivity

We consider the two-dimensional advection-diffusion equation on a bounded domain subject to either Dirichlet or von Neumann boundary conditions and study both time-independent and time-periodic cases involving Liouville integrable Hamiltonians that satisfy conditions conducive to applying the averaging principle. Transformation to action-angle coordinates permits averaging in time and angle, leading to an underlying eigenvalue equation that allows for separation of the angle and action coordinates. The result is a one-dimensional second-order equation involving an anti-symmetric imaginary potential. For radial flows on a disk or an annulus, we rigorously apply existing complex-plane WKBJ methods to study the spectral properties in the semi-classical limit for vanishing diffusivity. In this limit, the spectrum is found to be a complicated set consisting of lines related to Stokes graphs. Eigenvalues in the neighborhood of these graphs exhibit nonlinear scaling with respect to diffusivity leading to convection-enhanced rates of dissipation (relaxation, mixing) for initial data which are mean-free in the angle coordinate. These branches coexist with a diffusive branch of eigenvalues that scale linearly with diffusivity and contain the principal eigenvalue (no dissipation enhancement).

physics.flu-dyn↗
arXiv2013

The space-fractional diffusion-advection equation: Analytical solutions and critical assessment of numerical solutions

The present work provides a critical assessment of numerical solutions of the space-fractional diffusion-advection equation, which is of high significance for applications in various natural sciences. In view of the fact that, in contrast to the case of normal (Gaussian) diffusion, no standard methods and corresponding numerical codes for anomalous diffusion problems have been established yet, it is of importance to critically assess the accuracy and performance of existing approaches. Three numerical methods, namely a finite-difference method, the so-called matrix transfer technique, and a Monte-Carlo method based on the solution of stochastic differential equations, are analyzed and compared by applying them to three selected test problems for which analytical or semi-analytical solutions were known or are newly derived. The accuracy and performance differences are critically discussed with the result that the use of stochastic differential equations appears to be advantageous.

cond-mat.stat-mech↗
arXiv2013

Propagation of ultra-short solitons in stochastic Maxwell's equations

We study the propagation of ultra-short short solitons in a cubic nonlinear medium modeled by nonlinear Maxwell's equations with stochastic variations of media. We consider three cases: variations of (a) the dispersion, (b) the phase velocity, (c) the nonlinear coefficient. Using a modified multi-scale expansion for stochastic systems, we derive new stochastic generalizations of the short pulse equation that approximate the solutions of stochastic nonlinear Maxwell's equations. Numerical simulations show that soliton solutions of the short pulse equation propagate stably in stochastic nonlinear Maxwell's equations and that the generalized stochastic short pulse equations approximate the solutions to the stochastic Maxwell's equations over the distances under consideration. This holds for both a pathwise comparison of the stochastic equations as well as for a comparison of the resulting probability densities.

nlin.PS↗
arXiv2012

Symmetrization of advection-diffusion operators

We present a new method to transform an expanded class of non-selfadjoint advection-diffusion operators into self-adjoint operators. The transform is based on a combination of a point transform and Lie transform in conjunction with an asymptotic expansion in terms of the diffusivity. We illustrate the method in the context of simple shear flow where the expansion is exact and all transformation steps can be performed explicitly.

physics.flu-dyn↗
arXiv2012

Higher-order corrections to the short-pulse equation

Using renormalization group techniques, we derive an extended short- pulse equation as approximation to a nonlinear wave equation. We investigate the new equation numerically and show that the new equation captures efficiently higher- order effects on pulse propagation in cubic nonlinear media. We illustrate our findings using one- and two-soliton solutions of the first-order short-pulse equation as initial conditions in the nonlinear wave equation.

nlin.SI↗
arXiv2009

Quickest detection in coupled systems

This work considers the problem of quickest detection of signals in a coupled system of N sensors, which receive continuous sequential observations from the environment. It is assumed that the signals, which are modeled a general Ito processes, are coupled across sensors, but that their onset times may differ from sensor to sensor. The objective is the optimal detection of the first time at which any sensor in the system receives a signal. The problem is formulated as a stochastic optimization problem in which an extended average Kullback- Leibler divergence criterion is used as a measure of detection delay, with a constraint on the mean time between false alarms. The case in which the sensors employ cumulative sum (CUSUM) strategies is considered, and it is proved that the minimum of N CUSUMs is asymptotically optimal as the mean time between false alarms increases without bound.

cs.IT↗
arXiv2008

Averaged dynamics of time-periodic advection diffusion equations in the limit of small diffusivity

We study the effect of advection and small diffusion on passive tracers. The advecting velocity field is assumed to have mean zero and to possess time-periodic stream lines. Using a canonical transform to action-angle variables followed by a Lie-transform, we derive an averaged equation describing the effective motion of the tracers. An estimate for the time validity of the first-order approximation is established. For particular cases of a regularized vortical flow we present explicit formulas for the coefficients of the averaged equation both at first and at second order. Numerical simulations indicate that the validity of the above first-order estimate extends to the second order.

physics.flu-dyn↗