Bayesian ODE solvers: the maximum a posteriori estimate
From MaRDI portal
Publication:2058726
DOI10.1007/S11222-021-09993-7zbMATH Open1475.62061arXiv2004.00623OpenAlexW3135278807MaRDI QIDQ2058726FDOQ2058726
Simo Särkkä, Filip Tronarp, Philipp Hennig
Publication date: 9 December 2021
Published in: Statistics and Computing (Search for Journal in Brave)
Abstract: It has recently been established that the numerical solution of ordinary differential equations can be posed as a nonlinear Bayesian inference problem, which can be approximately solved via Gaussian filtering and smoothing, whenever a Gauss--Markov prior is used. In this paper the class of times differentiable linear time invariant Gauss--Markov priors is considered. A taxonomy of Gaussian estimators is established, with the maximum a posteriori estimate at the top of the hierarchy, which can be computed with the iterated extended Kalman smoother. The remaining three classes are termed explicit, semi-implicit, and implicit, which are in similarity with the classical notions corresponding to conditions on the vector field, under which the filter update produces a local maximum a posteriori estimate. The maximum a posteriori estimate corresponds to an optimal interpolant in the reproducing Hilbert space associated with the prior, which in the present case is equivalent to a Sobolev space of smoothness . Consequently, using methods from scattered data approximation and nonlinear analysis in Sobolev spaces, it is shown that the maximum a posteriori estimate converges to the true solution at a polynomial rate in the fill-distance (maximum step size) subject to mild conditions on the vector field. The methodology developed provides a novel and more natural approach to study the convergence of these estimators than classical methods of convergence analysis. The methods and theoretical results are demonstrated in numerical examples.
Full work available at URL: https://arxiv.org/abs/2004.00623
Recommendations
- Convergence rates of Gaussian ODE filters
- A probabilistic model for the numerical solution of initial value problems
- Probabilistic solutions to ordinary differential equations as nonlinear Bayesian filtering: a new perspective
- scientific article; zbMATH DE number 802686
- Bayesian analysis of ODEs: solver optimal accuracy and Bayes factors
Cites Work
- Title not available (Why is that?)
- Distributed Optimization and Statistical Learning via the Alternating Direction Method of Multipliers
- Gaussian processes for machine learning.
- Some results on Tchebycheffian spline functions and stochastic processes
- Mathematical Foundations of Infinite-Dimensional Statistical Models
- Computing integrals involving the matrix exponential
- Probabilistic numerics and uncertainty in computations
- A Correspondence Between Bayesian Estimation on Stochastic Processes and Smoothing by Splines
- Title not available (Why is that?)
- Stochastic differential equations. An introduction with applications.
- Title not available (Why is that?)
- Asymptotic analysis of penalized likelihood and related estimators
- Title not available (Why is that?)
- Numerical Methods for Ordinary Differential Equations
- The theory of differential equations. Classical and qualitative
- Title not available (Why is that?)
- Title not available (Why is that?)
- Title not available (Why is that?)
- An extension of a bound for functions in Sobolev spaces, with applications to \((m, s)\)-spline interpolation and smoothing
- Pushing the Limits of Contemporary Statistics: Contributions in Honor of Jayanta K. Ghosh
- The Iterated Kalman Smoother as a Gauss–Newton Method
- Bayesian filtering and smoothing
- Title not available (Why is that?)
- Bayesian solution uncertainty quantification for differential equations
- Maximum Likelihood Estimation and Uncertainty Quantification for Gaussian Process Approximation of Deterministic Functions
- A globalization scheme for the generalized Gauss-Newton method
- A property of multiplication in Sobolev spaces. Some applications
- Nemitsky operators on Sobolev spaces
- Stochastic interpretations and recursive algorithms for spline functions
- Convergence analysis of deterministic kernel-based quadrature rules in misspecified settings
- Probabilistic solutions to ordinary differential equations as nonlinear Bayesian filtering: a new perspective
- A modern retrospective on probabilistic numerics
- Bayesian Probabilistic Numerical Methods
- Error Bounds for Polynomial Spline Interpolation
- Statistical analysis of differential equations: introducing probability measures on numerical solutions
- Random time step probabilistic methods for uncertainty quantification in chaotic and geometric numerical integration
- Vector-valued Lg-splines. I: Interpolating splines
- A class of approximate solutions to linear operator equations
- Discrete-Time Solutions to the Continuous-Time Differential Lyapunov Equation With Applications to Kalman Filtering
- Gaussian kernel quadrature at scaled Gauss-Hermite nodes
- A role for symmetry in the Bayesian solution of differential equations
- Convergence rates of Gaussian ODE filters
- Strong convergence rates of probabilistic integrators for ordinary differential equations
- A probabilistic model for the numerical solution of initial value problems
- Title not available (Why is that?)
- Applied Stochastic Differential Equations
- Iterated Extended Kalman Smoother-Based Variable Splitting for $L_1$-Regularized State Estimation
- Estimation of Ordinary Differential Equation Models with Discretization Error Quantification
Cited In (4)
- Bayesian numerical methods for nonlinear partial differential equations
- Modelling the discretization error of initial value problems using the Wishart distribution
- A role for symmetry in the Bayesian solution of differential equations
- Bayesian solution uncertainty quantification for differential equations
This page was built for publication: Bayesian ODE solvers: the maximum a posteriori estimate
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2058726)
