Séminaire de mathématiques appliquées

Dans cette présentation, nous nous intéresserons à deux méthodes numériques indépen- dantes permettant d’accélérer et d’améliorer la résolution des EDP.

Récemment, ParaOpt, un algorithme parallèle en temps a été proposé pour résoudre les systèmes découlant de problèmes de contrôle optimal associés à des équations différentielles partielles (EDP). Cet algorithme combine la résolution des problèmes d’évolution forward/backward avec la boucle d’optimisation. Sa convergence et la précision de l’approximation numérique dépendent du schéma de pas de temps utilisé pour discrétiser le problème. Une première analyse de convergence a été présentée dans M.J. Gander, F. Kwok et J. Salomon SISC (2020) dans le cas restreint de la méthode implicite d’Euler pour les problèmes de contrôle optimal linéaire-quadratique impliquant des systèmes dissipatifs. Dans cette présentation, nous généraliserons ce résultat au cas où une méthode de Runge-Kutta est utilisée pour discrétiser le problème de contrôle optimal. Nous expliquons en particulier comment la discrétisation de la fonctionnelle doit être adaptée au schéma considéré et comment des conditions supplémentaires s’ajoutent aux conditions d’ordre habituel des méthodes de Runge-Kutta pour obtenir l’ordre de convergence attendu. Aphynity [Yuan Yin et al., J. Stat. Mech. (2021) 124012] est une approche consistant à ajouter un terme correctif sous la forme d’un réseau neurone à une EDP afin de prévoir avec précision l’évolution d’un système et d’identifier correctement ses paramètres physiques. L’apprentissage est alors effectué dans la boucle externe du solveur considéré, de sorte que l’entraînement se fait indirectement par le biais d’un schéma en temps. Nous décrirons l’influence du schéma choisi sur le réseau résultant et montrerons, dans des cas particuliers, que l’ordre d’approximation des termes corrigés correspond à l’ordre du schéma utilisé. Nous expliquerons également comment optimiser l’apprentissage par une approche de type Richardson.

Bayesian Inversion consists of deriving the posterior distribution of unknown parameters or functions from partial and indirect observations of a system. When the dimension of the search space is high or infinite, methods leveraging local information, such as derivatives of different orders, of the target probability measure have the advantages to converge faster than Monte-Carlo sampling techniques. Nevertheless, many applications are characterized by posterior distributions with low regularity or gradients that are intractable to compute. An interesting research direction consists in using interacting particle systems to explore the potential landscape, and Ensemble Kalman Sampler (EKS) is one of those. In this talk, we consider a simplified EKS dynamics, where the gradient of the potential is approximated by finite differences using independent Ornstein-Uhlenbeck processes that explore the neighborhood of the candidate parameter. We will characterize the invariant distribution of this system and compare its dynamics to the overdamped Langevin process.

In this work, we introduce a novel collocation-based Model Order Reduction strategy, called collocation Model Order Reduction (cMOR), proposed as an alternative to classical projection-based Model Order Reduction (pMOR). Unlike pMOR, which relies on Galerkin or Petrov–Galerkin projection onto a reduced space, cMOR determines the reduced solution by solving the governing equations at a strategically selected set of collocation points, identified through hyper-reduction techniques. The method preserves the typical two-stage structure of MOR: an offline phase, where a Reduced Basis (RB) is constructed from snapshots, and an online phase, where the High-Dimensional Model (HDM) is evaluated only on the empirical subset of points and the full solution is reconstructed via the RB.

The cMOR framework is analyzed theoretically in terms of stability and convergence, and it is applied within both linear reduced spaces and nonlinear approximation manifolds (NAM), the latter enhancing the ability to represent complex multiscale structures by embedding the solution on nonlinear manifolds rather than linear subspaces. Moreover, the method is implemented using advanced numerical technologies, including ADER schemes on unsteady Chimera grids, enabling efficient treatment of convection-dominated PDEs on evolving geometries. These applications highlight the robustness of cMOR in challenging scenarios—such as advection-dominated dynamics, moving multi-block grids, and high-dimensional parametric variations—where classical pMOR often suffers from instability or reduced accuracy due to the Kolmogorov N-width barrier.

Beyond traditional discretization frameworks, cMOR also demonstrates strong compatibility with varied computational settings, from numerical schemes to grid-based solvers and neural-network-assisted reduced models, thanks to its collocation-driven formulation. Numerical experiments across these diverse contexts confirm that cMOR achieves substantial computational savings while maintaining high accuracy, ensuring straightforward integration into existing simulation workflows and machine-learning-augmented pipelines.

TBA

State-Space Models (SSMs) are deterministic or stochastic dynamical systems defined by two processes. The state process, which is not observed directly, models the transformation of the states over time. On another hand, the observation process produces the observables on which model fitting and prediction are based. Ecology frequently uses stochastic SSMs to represent the imperfectly observed dynamics of population sizes or animal movement. However, several simulation-based evaluations of model performance suggest broad identifiability issues in ecological SSMs. Formal SSM identifiability is typically investigated using exhaustive summaries, which are simplified representations of the model. The theory on exhaustive summaries is largely based on continuous-time deterministic modelling and those for ecological SSMs have developed by analogy. While the discreteness of time does not constitute a challenge, finding a good exhaustive summary for a stochastic SSM is more difficult. The strategy adopted so far has been to create exhaustive summaries based on a transfer function of the expectations of the stochastic process. However, this evaluation of identifiability does not allow to take into account the possible dependency between the variance parameters and the process parameters. We show that the output spectral density plays a key role in stochastic SSM identifiability assessment. This allows us to define a new suitable exhaustive summary. Using several ecological examples, we show that usual ecological models are often theoretically identifiable, suggesting that most SSM estimation problems are due to practical rather than theoretical identifiability issues.

Spline interpolation is a widely used class of methods for solving interpolation problems by constructing smooth interpolants that minimize a regularized energy functional involving the Laplacian operator. While many existing approaches focus on Euclidean domains or the sphere, relying on the spectral properties of the Laplacian, this work introduces a method for spline interpolation on general manifolds by exploiting its equivalence with kriging. Specifically, the proposed approach uses finite element approximations of random fields defined over the manifold, based on Gaussian Markov Random Fields and a discretization of the Laplace-Beltrami operator on a triangulated mesh. This framework enables the modeling of spatial fields with local anisotropies through domain deformation. The method is first validated on the sphere using both analytical test cases and a pollution-related study, and is compared to the classical spherical harmonics-based method. Additional experiments on the surface of a cylinder further illustrate the generality of the approach.

TBA