-
Sampling Methods for Bayesian Inference Involving Convergent Noisy Approximations of Forward Maps
Authors:
Giacomo Garegnani
Abstract:
We present Bayesian techniques for solving inverse problems which involve mean-square convergent random approximations of the forward map. Noisy approximations of the forward map arise in several fields, such as multiscale problems and probabilistic numerical methods. In these fields, a random approximation can enhance the quality or the efficiency of the inference procedure, but entails additiona…
▽ More
We present Bayesian techniques for solving inverse problems which involve mean-square convergent random approximations of the forward map. Noisy approximations of the forward map arise in several fields, such as multiscale problems and probabilistic numerical methods. In these fields, a random approximation can enhance the quality or the efficiency of the inference procedure, but entails additional theoretical and computational difficulties due to the randomness of the forward map. A standard technique to address this issue is to combine Monte Carlo averaging with Markov chain Monte Carlo samplers, as for example in the pseudo-marginal Metropolis--Hastings methods. In this paper, we consider mean-square convergent random approximations, and quantify how Monte Carlo errors propagate from the forward map to the solution of the inverse problems. Moreover, we review and describe simple techniques to solve such inverse problems, and compare performances with a series of numerical experiments.
△ Less
Submitted 5 November, 2021;
originally announced November 2021.
-
Robust Estimation of Effective Diffusions from Multiscale Data
Authors:
Giacomo Garegnani,
Andrea Zanoni
Abstract:
We present a novel methodology based on filtered data and moving averages for estimating effective dynamics from observations of multiscale systems. We show in a semi-parametric framework of the Langevin type that our approach is asymptotically unbiased with respect to the theory of homogenization. Moreover, we demonstrate on a range of challenging numerical experiments that our method is accurate…
▽ More
We present a novel methodology based on filtered data and moving averages for estimating effective dynamics from observations of multiscale systems. We show in a semi-parametric framework of the Langevin type that our approach is asymptotically unbiased with respect to the theory of homogenization. Moreover, we demonstrate on a range of challenging numerical experiments that our method is accurate in extracting coarse-grained dynamics from multiscale data. In particular, the estimators we propose are more robust and require less knowledge of the full model than the standard technique of subsampling, which is widely employed in practice in this setting.
△ Less
Submitted 24 January, 2022; v1 submitted 7 September, 2021;
originally announced September 2021.
-
A probabilistic finite element method based on random meshes: Error estimators and Bayesian inverse problems
Authors:
Assyr Abdulle,
Giacomo Garegnani
Abstract:
We present a novel probabilistic finite element method (FEM) for the solution and uncertainty quantification of elliptic partial differential equations based on random meshes, which we call random mesh FEM (RM-FEM). Our methodology allows to introduce a probability measure on standard piecewise linear FEM. We present a posteriori error estimators based uniquely on probabilistic information. A seri…
▽ More
We present a novel probabilistic finite element method (FEM) for the solution and uncertainty quantification of elliptic partial differential equations based on random meshes, which we call random mesh FEM (RM-FEM). Our methodology allows to introduce a probability measure on standard piecewise linear FEM. We present a posteriori error estimators based uniquely on probabilistic information. A series of numerical experiments illustrates the potential of the RM-FEM for error estimation and validates our analysis. We furthermore demonstrate how employing the RM-FEM enhances the quality of the solution of Bayesian inverse problems, thus allowing a better quantification of numerical errors in pipelines of computations.
△ Less
Submitted 10 March, 2021;
originally announced March 2021.
-
Drift Estimation of Multiscale Diffusions Based on Filtered Data
Authors:
Assyr Abdulle,
Giacomo Garegnani,
Grigorios A. Pavliotis,
Andrew M. Stuart,
Andrea Zanoni
Abstract:
We study the problem of drift estimation for two-scale continuous time series. We set ourselves in the framework of overdamped Langevin equations, for which a single-scale surrogate homogenized equation exists. In this setting, estimating the drift coefficient of the homogenized equation requires pre-processing of the data, often in the form of subsampling; this is because the two-scale equation a…
▽ More
We study the problem of drift estimation for two-scale continuous time series. We set ourselves in the framework of overdamped Langevin equations, for which a single-scale surrogate homogenized equation exists. In this setting, estimating the drift coefficient of the homogenized equation requires pre-processing of the data, often in the form of subsampling; this is because the two-scale equation and the homogenized single-scale equation are incompatible at small scales, generating mutually singular measures on the path space. We avoid subsampling and work instead with filtered data, found by application of an appropriate kernel function, and compute maximum likelihood estimators based on the filtered process. We show that the estimators we propose are asymptotically unbiased and demonstrate numerically the advantages of our method with respect to subsampling. Finally, we show how our filtered data methodology can be combined with Bayesian techniques and provide a full uncertainty quantification of the inference procedure.
△ Less
Submitted 6 June, 2021; v1 submitted 28 September, 2020;
originally announced September 2020.
-
Integrated approach for the identification of spatial patterns related to renewable energy potential in European territories
Authors:
Chiara Scaramuzzino,
Giulia Garegnani,
Pietro Zambelli
Abstract:
The study presents an effort to classify the territories of a specific area, according to similarities in the estimated potential of their renewable sources, considering also their economic and sociodemographic structure and their geographic features. Specifically, the paper focuses on the area of EU28 and Switzerland and uses as basis for the analysis, data estimating the potential of renewable e…
▽ More
The study presents an effort to classify the territories of a specific area, according to similarities in the estimated potential of their renewable sources, considering also their economic and sociodemographic structure and their geographic features. Specifically, the paper focuses on the area of EU28 and Switzerland and uses as basis for the analysis, data estimating the potential of renewable energy sources collected and elaborated in the framework of the project HotMaps (Horizon 2020). The method used to group the territorial units is cluster analysis, and specifically the k-means algorithm. The data present some interesting patterns and the territories of EU28 and Switzerland at NUTS3 level are classified into 17 clusters. The analysis shows the presence of heterogeneity within national borders and among territories comprised in the macro regions target of specific EU programmes, specifically the Adriatic-Ionian region, the Alpine region, the Baltic Sea region and the Danube region. The results of this research are meant to be used by European policy makers in developing more focused transnational renewable energy policies and strategies.
△ Less
Submitted 12 November, 2019; v1 submitted 31 October, 2019;
originally announced November 2019.
-
Ensemble Kalman filter for multiscale inverse problems
Authors:
Assyr Abdulle,
Giacomo Garegnani,
Andrea Zanoni
Abstract:
We present a novel algorithm based on the ensemble Kalman filter to solve inverse problems involving multiscale elliptic partial differential equations. Our method is based on numerical homogenization and finite element discretization and allows to recover a highly oscillatory tensor from measurements of the multiscale solution in a computationally inexpensive manner. The properties of the approxi…
▽ More
We present a novel algorithm based on the ensemble Kalman filter to solve inverse problems involving multiscale elliptic partial differential equations. Our method is based on numerical homogenization and finite element discretization and allows to recover a highly oscillatory tensor from measurements of the multiscale solution in a computationally inexpensive manner. The properties of the approximate solution are analysed with respect to the multiscale and discretization parameters, and a convergence result is shown to hold. A reinterpretation of the solution from a Bayesian perspective is provided, and convergence of the approximate conditional posterior distribution is proved with respect to the Wasserstein distance. A numerical experiment validates our methodology, with a particular emphasis on modelling error and computational cost.
△ Less
Submitted 25 June, 2020; v1 submitted 15 August, 2019;
originally announced August 2019.
-
Random time step probabilistic methods for uncertainty quantification in chaotic and geometric numerical integration
Authors:
Assyr Abdulle,
Giacomo Garegnani
Abstract:
A novel probabilistic numerical method for quantifying the uncertainty induced by the time integration of ordinary differential equations (ODEs) is introduced. Departing from the classical strategy to randomize ODE solvers by adding a random forcing term, we show that a probability measure over the numerical solution of ODEs can be obtained by introducing suitable random time-steps in a classical…
▽ More
A novel probabilistic numerical method for quantifying the uncertainty induced by the time integration of ordinary differential equations (ODEs) is introduced. Departing from the classical strategy to randomize ODE solvers by adding a random forcing term, we show that a probability measure over the numerical solution of ODEs can be obtained by introducing suitable random time-steps in a classical time integrator. This intrinsic randomization allows for the conservation of geometric properties of the underlying deterministic integrator such as mass conservation, symplecticity or conservation of first integrals. Weak and mean-square convergence analysis are derived. We also analyse the convergence of the Monte Carlo estimator for the proposed random time step method and show that the measure obtained with repeated sampling converges in the mean-square sense independently of the number of samples. Numerical examples including chaotic Hamiltonian systems, chemical reactions and Bayesian inferential problems illustrate the accuracy, robustness and versatility of our probabilistic numerical method.
△ Less
Submitted 21 January, 2020; v1 submitted 4 January, 2018;
originally announced January 2018.
