arXiv subjects

Alexandre B. Simas

Publications and source records attributed to Alexandre B. Simas.

At least 19 recordsLinked to original sources

arXiv2026

Mass Lumping and Numerical Quadrature for Approximation of Fractional Elliptic Differential Equations Driven by Gaussian White Noise

Fractional elliptic stochastic partial differential equations (SPDEs) are widely used in statistics and machine learning for computationally efficient and flexible modeling of Gaussian random fields. The computational efficiency of the SPDE approach relies on finite element approximations combined with numerical quadrature and mass lumping, which enable sparse matrix methods during inference. Although many works have studied finite element approximations of fractional SPDEs, the effect of the mass lumping and quadrature approximations used in practice has not been fully analyzed. To fill this gap, we derive convergence rates for numerical approximations of fractional SPDEs based on finite element discretizations combined with numerical quadrature and mass lumping. Specifically, we obtain explicit convergence rates for the mean-squared error of the covariance function in a general framework that covers the main settings where mass lumping is used in the SPDE approach. We also analyze non-stationary variance-control factors of the form $L^\beta(\tau u)=\mathcal{W}$, where $\tau$ is spatially varying, and derive covariance error estimates showing how the regularity of $\tau$ affects the convergence rate. As specific examples, we provide results for random fields on bounded Euclidean domains, Riemannian manifolds, and metric graphs. Numerical experiments are presented that confirm the theoretical results.

math.NA↗

arXiv2026

Numerical approximation of fractional diffusion equations on metric graphs

We study fractional diffusion equations on compact metric graphs, where the nonlocal dynamics is governed by fractional powers of the shifted Kirchhoff-Laplacian. Building on recent advances in the analysis of fractional operators on metric graphs, we establish a rigorous mathematical framework and propose a fully discrete scheme based on backward Euler time-stepping and finite element discretization. To approximate the action of the fractional operator, we employ rational approximations, reducing the problem to a sequence of sparse elliptic solves for efficient implementation. We derive error estimates for the temporal, spatial, and rational discretizations, and confirm convergence through numerical experiments.

math.NA↗

arXiv2026

Finite element and box-method discretizations for fractional elliptic problems with quadrature and mass lumping

We analyze numerical approximation of the fractional elliptic problem $L^{\beta}u=f$, ${\beta>0}$, where $L$ is a second-order self-adjoint elliptic operator with homogeneous Dirichlet or Neumann boundary conditions. The paper develops a unified conforming piecewise linear framework that covers both the standard finite element discretization and the box-method discretization of fractional powers. The key point is that the discrete fractional operator is defined with respect to an admissible inner product on the trial space. This includes, in particular, the standard $L^{2}$ inner product and the quadrature-based mass-lumped inner product, and we also identify a broader family of admissible inner products interpolating between these two realizations. Within this framework, we show that the mass-lumped choice yields the intrinsic fractional box discretization, namely the one obtained by taking fractional powers of the nonfractional box solution operator. For both the finite element and box-method realizations, we establish error estimates under natural consistency assumptions, making explicit the effect of load quadrature in the box case. The analysis applies directly to practical schemes and is supported by numerical experiments in one and two space dimensions.

math.NA↗

arXiv2026

A Unified and Computationally Efficient Non-Gaussian Statistical Modeling Framework

Datasets that exhibit non-Gaussian characteristics are common in many fields, while the current modeling framework and available software for non-Gaussian models is limited. We introduce Linear Latent Non-Gaussian Models (LLnGMs), a unified and computationally efficient statistical modeling framework that extends a class of latent Gaussian models to allow for latent non-Gaussian processes. The framework unifies several popular models, from simple temporal models to complex spatial-temporal and multivariate models, facilitating natural non-Gaussian extensions. Computationally efficient Bayesian inference, with theoretical guarantees, is developed based on stochastic gradient descent estimation. The R package \texttt{ngme2}, which implements the framework, is presented and demonstrated through a wide range of applications including novel non-Gaussian spatial and spatio-temporal models.

stat.ME↗

arXiv2026

Geometric ergodicity of Gibbs samplers for linear latent models with GIG variance mixtures

We study geometric ergodicity of the Gibbs sampler for linear latent non-Gaussian models (LLnGMs), a class of hierarchical models in which conditional Gaussian structure is preserved through generalized inverse Gaussian (GIG) variance-mixture augmentation. Two complementary routes to geometric ergodicity are developed for the marginal chain on the mixing variables. First, we show that the associated Markov operator is trace-class, and hence admits a spectral gap, over a large portion of the GIG parameter space. Second, for the remaining boundary and heavy-tail regimes, we establish geometric ergodicity via drift and minorization, subject to an explicit null-smallness condition that quantifies how the drift interacts with the null space of the observation operator. Together, these results cover the full GIG parameter space, including the normal-inverse Gaussian, generalized asymmetric Laplace, and Student-$t$ special cases. The geometric ergodicity of this chain underpins the consistency of Gibbs-based stochastic-gradient estimators for maximum likelihood estimation, and we provide conditions that make the required integrability checks transparent. Numerical experiments illustrate the theoretical findings, contrasting mixing efficiency across parameter regimes and probing the role of the null-smallness constant.

math.ST↗

arXiv2025

Forecasting the Term Structure of Interest Rates with SPDE-Based Models

The Dynamic Nelson--Siegel (DNS) model is a widely used framework for term structure forecasting. We propose a novel extension that models DNS residuals as a Gaussian random field, capturing dependence across both time and maturity. The residual field is represented via a stochastic partial differential equation (SPDE), enabling flexible covariance structures and scalable Bayesian inference through sparse precision matrices. We consider a range of SPDE specifications, including stationary, non-stationary, anisotropic, and nonseparable models. The SPDE--DNS model is estimated in a Bayesian framework using the integrated nested Laplace approximation (INLA), jointly inferring latent DNS factors and the residual field. Empirical results show that the SPDE-based extensions improve both point and probabilistic forecasts relative to standard benchmarks. When applied in a mean--variance bond portfolio framework, the forecasts generate economically meaningful utility gains, measured as performance fees relative to a Bayesian DNS benchmark under monthly rebalancing. Importantly, incorporating the structured SPDE residual substantially reduces cross-maturity and intertemporal dependence in the remaining measurement error, bringing it closer to white noise. These findings highlight the advantages of combining DNS with SPDE-driven residual modeling for flexible, interpretable, and computationally efficient yield curve forecasting.

stat.AP↗

arXiv2025

Fractional and Integer Order Sobolev Spaces for Compact Metric Graphs

Given a compact metric graph $\Gamma$ and the Laplacian $\Delta_{\Gamma}$ coupled with standard (Kirchhoff) vertex conditions, solutions to fractional elliptic partial differential equations of the form $(\kappa^2 - \Delta_{\Gamma})^{\alpha/2}u=f$ on $\Gamma$ exhibit a distinctive regularity structure: even-order derivatives are continuous across vertices, while odd-order derivatives may be discontinuous. This non-standard smoothness property precludes the direct application of classical tools from real functional analysis. Because of this, we introduce and systematically study new families of Sobolev spaces tailored to this setting. We define these spaces, denoted $W^{\alpha,p}(\Gamma)$ and $H^{\alpha}(\Gamma)$, to respect the continuity constraints on even-order derivatives at vertices, while permitting discontinuities in odd-order derivatives. We establish their fundamental properties, including characterizations, embedding theorems into H\"older and Lebesgue spaces, and compactness results. A central contribution in this investigation is the derivation of uniform bounds on the supremum norm of eigenfunctions for a class of Laplacians on metric graphs, a result of independent interest. Finally, we demonstrate that these spaces provide a natural framework for analyzing the regularity of solutions to fractional elliptic PDEs and SPDEs driven by Gaussian white noise on metric graphs, in particular, establishing a general characterization of the domain of the fractional powers of $(\kappa^2-\Delta_{\Gamma})$ and $(\kappa^2-\nabla(a\nabla))$ in terms of the Sobolev spaces we introduce, thereby extending all previously known characterizations in the literature, and improving the regularity results previously obtained to their sharp counterparts (with general fractional powers). We also show that these spaces are fundamental to the characterization of Gaussian free fields on metric graphs.

math.AP↗

arXiv2025

Spectral Theory of Krein-Feller Type Operators and Applications in Stochastic Fractional Elliptic and Parabolic Equations

It has been shown that the space $C^{\infty}_{W,V}(\mathbb{T})$, introduced in Simas and Sousa (Potential Analysis, 2025), is the natural regularity space for solutions of the eigenvalue problem $\Delta_{W,V} u = \lambda u$ on the torus $\mathbb{T}$, where $\Delta_{W,V} = \frac{d^{+}}{dV}\frac{d^{-}}{dW}$ is the Krein Feller operator in the case where $W$ and $V$ are strictly increasing and right continuous (respectively left continuous), possibly with dense sets of discontinuities. In this work we provide conditions ensuring that every function in $C^{\infty}_{W,V}(\mathbb{T})$, which may be highly discontinuous, admits a series expansion that generalizes the classical Taylor expansion. A central feature of our approach is that all proofs are nonstandard, since classical analytical and spectral arguments cannot be adapted to this singular setting. Using these methods we characterize the eigenvectors of $\Delta_{W,V}$ in terms of generalized trigonometric functions and obtain an asymptotic lower bound for the associated eigenvalues. We also derive a sharp upper bound for the convergence exponent of these eigenvalues, and as a consequence we prove that $C^{\infty}_{W,V}(\mathbb{T})$ is a nuclear space. Further consequences include results on the asymptotic behavior of eigenvalues of compact operators and improvements in traceability. As a final application we establish existence results for generalized fractional stochastic and deterministic differential equations, as well as for parabolic stochastic partial differential equations acting on nuclear spaces.

math.AP↗

arXiv2025

Incorporating Correlated Nugget Effects in Multivariate Spatial Models: An Application to Argo Ocean Data

Accurate analysis of global oceanographic data, such as temperature and salinity profiles from the Argo program, requires geostatistical models capable of capturing complex spatial dependencies. This study introduces Gaussian and non-Gaussian hierarchical multivariate Mat\'ern-SPDE models with correlated nugget effects to account for small-scale variability and measurement error correlations. Using simulations and Argo data, we demonstrate that incorporating correlated nugget effects significantly improves the accuracy of parameter estimation and spatial prediction in both Gaussian and non-Gaussian multivariate spatial processes. When applied to global ocean temperature and salinity data, our model yields lower correlation estimates between fields compared to models that assume independent noise. This suggests that traditional models may overestimate the underlying field correlation. By separating these effects, our approach captures fine-scale oceanic patterns more effectively. These findings show the importance of relaxing the assumption of independent measurement errors in multivariate hierarchical models.

stat.ME↗

arXiv2025

rSPDE: tools for statistical modeling using fractional SPDEs

The R software package rSPDE contains methods for approximating Gaussian random fields based on fractional-order stochastic partial differential equations (SPDEs). A common example of such fields are Whittle-Mat\'ern fields on bounded domains in $\mathbb{R}^d$, manifolds, or metric graphs. The package also implements various other models which are briefly introduced in this article. Besides the approximation methods, the package contains methods for simulation, prediction, and statistical inference for such models, as well as interfaces to INLA, inlabru and MetricGraph. With these interfaces, fractional-order SPDEs can be used as model components in general latent Gaussian models, for which full Bayesian inference can be performed, also for fractional models on metric graphs. This includes estimation of the smoothness parameter of the fields. This article describes the computational methods used in the package and summarizes the theoretical basis for these. The main functions of the package are introduced, and their usage is illustrated through various examples.

stat.CO↗

arXiv2025

One-sided measure theoretic elliptic operators and applications to SDEs driven by Gaussian white noise with atomic intensity

We define the operator $D^+_VD^-_W:=\Delta_{W,V}$ on the one-dimensional torus $\mathbb{T}$. Here, $W$ and $V$ are functions inducing (possibly atomic) positive Borel measures on $\mathbb{T}$, and the derivatives are generalized lateral derivatives. For the first time in this work, the space of test functions $C^{\infty}_{W,V}(\mathbb{T})$ emerges as the natural regularity space for solutions of the eigenproblem associated with $\Delta_{W,V}$. Moreover, these spaces are essential for characterizing the energetic space $H_{W,V}(\mathbb{T})$ as a Sobolev-type space. By observing that the Sobolev-type spaces $H_{W,V}(\mathbb{T})$ with additional Dirichlet conditions are reproducing kernel Hilbert spaces, we introduce the so-called $W$-Brownian bridges as mean-zero Gaussian processes with associated Cameron-Martin spaces derived from these spaces. This framework allows us to introduce $W$-Brownian motion as a Feller process with a two-parameter semigroup and c\`adl\`ag sample paths, whose jumps are subordinated to the jumps of $W$. We establish a deep connection between $W$-Brownian motion and these Sobolev-type spaces through their associated Cameron-Martin spaces. Finally, as applications of the developed theory, we demonstrate the existence and uniqueness of related deterministic and stochastic differential equations.

math.AP↗

arXiv2025

Log-Gaussian Cox Processes on General Metric Graphs

The modeling of spatial point processes has advanced considerably, yet extending these models to non-Euclidean domains, such as road networks, remains a challenging problem. We propose a novel framework for log-Gaussian Cox processes on general compact metric graphs by leveraging the Gaussian Whittle-Mat\'ern fields, which are solutions to fractional-order stochastic differential equations on metric graphs. To achieve computationally efficient likelihood-based inference, we introduce a numerical approximation of the likelihood that eliminates the need to approximate the Gaussian process. This method, coupled with the exact evaluation of finite-dimensional distributions for Whittle-Mat\'ern fields with integer smoothness, ensures scalability and theoretical rigour, with derived convergence rates for posterior distributions. The framework is implemented in the open-source MetricGraph R package, which integrates seamlessly with R-INLA to support fully Bayesian inference. We demonstrate the applicability and scalability of this approach through an analysis of road accident data from Al-Ahsa, Saudi Arabia, consisting of over 150,000 road segments. By identifying high-risk road segments using exceedance probabilities and excursion sets, our framework provides localized insights into accident hotspots and offers a powerful tool for modeling spatial point processes directly on complex networks.

stat.ME↗

arXiv2025

A new class of non-stationary Gaussian fields with general smoothness on metric graphs

The increasing availability of network data has driven the development of advanced statistical models specifically designed for metric graphs, where Gaussian processes play a pivotal role. While models such as Whittle-Mat\'ern fields have been introduced, there remains a lack of practically applicable options that accommodate flexible non-stationary covariance structures or general smoothness. To address this gap, we propose a novel class of generalized Whittle-Mat\'ern fields, which are rigorously defined on general compact metric graphs and permit both non-stationarity and arbitrary smoothness. We establish new regularity results for these fields, which extend even to the standard Whittle-Mat\'ern case. Furthermore, we introduce a method to approximate the covariance operator of these processes by combining the finite element method with a rational approximation of the operator's fractional power, enabling computationally efficient Bayesian inference for large datasets. Theoretical guarantees are provided by deriving explicit convergence rates for the covariance approximation error, and the practical utility of our approach is demonstrated through simulation studies and an application to traffic speed data, highlighting the flexibility and effectiveness of the proposed model class.

stat.ME↗

arXiv2025

An explicit link between graphical models and Gaussian Markov random fields on metric graphs

We derive an explicit link between Gaussian Markov random fields on metric graphs and graphical models, and in particular show that a Markov random field restricted to the vertices of the graph is, under mild regularity conditions, a Gaussian graphical model with a distribution which is faithful to its pairwise independence graph, which coincides with the neighbor structure of the metric graph. This is used to show that there are no Gaussian random fields on general metric graphs which are both Markov and isotropic in some suitably regular metric on the graph, such as the geodesic or resistance metrics.

math.PR↗

arXiv2024

Linear cost and exponentially convergent approximation of Gaussian Mat\'ern processes on intervals

The computational cost for inference and prediction of statistical models based on Gaussian processes with Mat\'ern covariance functions scales cubicly with the number of observations, limiting their applicability to large data sets. The cost can be reduced in certain special cases, but there are currently no generally applicable exact methods with linear cost. Several approximate methods have been introduced to reduce the cost, but most of these lack theoretical guarantees for the accuracy. We consider Gaussian processes on bounded intervals with Mat\'ern covariance functions and for the first time develop a generally applicable method with linear cost and with a covariance error that decreases exponentially fast in the order $m$ of the proposed approximation. The method is based on an optimal rational approximation of the spectral density and results in an approximation that can be represented as a sum of $m$ independent Gaussian Markov processes, which facilitates easy usage in general software for statistical inference, enabling its efficient implementation in general statistical inference software packages. Besides the theoretical justifications, we demonstrate the accuracy empirically through carefully designed simulation studies which show that the method outperforms all state-of-the-art alternatives in terms of accuracy for a fixed computational cost in statistical tasks such as Gaussian process regression.

math.ST↗

arXiv2023

Wasserstein complexity penalization priors: a new class of penalizing complexity priors

Penalizing complexity (PC) priors provide a principled framework for reducing model complexity by penalizing the Kullback--Leibler Divergence (KLD) between a ``simple'' base model and a more complex model. However, constructing priors by penalizing the KLD becomes impossible in many cases because the KLD is infinite, and alternative principles often lose interpretability in terms of KLD. We propose a new class of priors, the Wasserstein complexity penalization (WCP) priors, which replace the KLD with the Wasserstein distance in the PC prior framework. WCP priors avoid the issue of infinite model distances and retain interpretability by adhering to adjusted principles. Additionally, we introduce the concept of base measures, removing the parameter dependency on the base model, and extend the framework to joint WCP priors for multiple parameters. These priors can be constructed analytically and we have both analytical and numerical implementations in R programming language. We demonstrate their use in previous PC prior applications and as well as new multivariate settings.

stat.ME↗

arXiv2023

Hydrodynamic limit of the multi-component slow boundary WASEP with collisions

In this article, we study the hydrodynamic limit for a stochastic interacting particle system whose dynamics consists in a superposition of several dynamics: the exclusion rule, that dictates that no more than a particle per site with a fixed velocity is allowed; a collision dynamics, that dictates that particles at the same site can collide and originate particles with new velocities such that the linear momentum is conserved; a boundary dynamics that injects and removes particle in the system. This last dynamics destroys the conservation law, and its strength is regulated by a parameter $\theta$. The goal is the derivation of the hydrodynamic limit, and the boundary conditions change drastically according to the value of $\theta$.

math.PR↗

arXiv2023

Markov properties of Gaussian random fields on compact metric graphs

There has recently been much interest in Gaussian fields on linear networks and, more generally, on compact metric graphs. One proposed strategy for defining such fields on a metric graph $\Gamma$ is through a covariance function that is isotropic in a metric on the graph. Another is through a fractional-order differential equation $L^{\alpha/2} (\tau u) = \mathcal{W}$ on $\Gamma$, where $L = \kappa^2 - \nabla(a\nabla)$ for (sufficiently nice) functions $\kappa, a$, and $\mathcal{W}$ is Gaussian white noise. We study Markov properties of these two types of fields. First, we show that no Gaussian random fields exist on general metric graphs that are both isotropic and Markov. Then, we show that the second type of fields, the generalized Whittle--Mat\'ern fields, are Markov if $\alpha\in\mathbb{N}$, and conversely, if $a$ and $\kappa$ are constant and $u$ is Markov, then $\alpha\in\mathbb{N}$. Further, if $\alpha\in\mathbb{N}$, a generalized Whittle--Mat\'ern field $u$ is Markov of order $\alpha$, which means that the field $u$ in one region $S\subset\Gamma$ is conditionally independent of $u$ in $\Gamma\setminus S$ given the values of $u$ and its $\alpha-1$ derivatives on $\partial S$. Finally, we provide two results as consequences of the theory developed: first we prove that the Markov property implies an explicit characterization of $u$ on a fixed edge $e$, revealing that the conditional distribution of $u$ on $e$ given the values at the two vertices connected to $e$ is independent of the geometry of $\Gamma$; second, we show that the solution to $L^{1/2}(\tau u) = \mathcal{W}$ on $\Gamma$ can obtained by conditioning independent generalized Whittle--Mat\'ern processes on the edges, with $\alpha=1$ and Neumann boundary conditions, on being continuous at the vertices.

math.PR↗