A probabilistic model for the numerical solution of initial value problems

DOI10.1007/S11222-017-9798-7zbMATH Open1505.62361arXiv1610.05261OpenAlexW2534928453MaRDI QIDQ2329752FDOQ2329752

Authors: Michael Schober, Simo Särkkä, Philipp Hennig 

Publication date: 18 October 2019

Published in: Statistics and Computing (Search for Journal in Brave)

Abstract: Like many numerical methods, solvers for initial value problems (IVPs) on ordinary differential equations estimate an analytically intractable quantity, using the results of tractable computations as inputs. This structure is closely connected to the notion of inference on latent variables in statistics. We describe a class of algorithms that formulate the solution to an IVP as inference on a latent path that is a draw from a Gaussian process probability measure (or equivalently, the solution of a linear stochastic differential equation). We then show that certain members of this class are connected precisely to generalized linear methods for ODEs, a number of Runge--Kutta methods, and Nordsieck methods. This probabilistic formulation of classic methods is valuable in two ways: analytically, it highlights implicit prior assumptions favoring certain approximate solutions to the IVP over others, and gives a precise meaning to the old observation that these methods act like filters. Practically, it endows the classic solvers with `docking points' for notions of uncertainty and prior information about the initial value, the value of the ODE itself, and the solution of the problem.

Full work available at URL: https://arxiv.org/abs/1610.05261

Recommendations

• Probabilistic solutions to ordinary differential equations as nonlinear Bayesian filtering: a new perspective
• Bayesian ODE solvers: the maximum a posteriori estimate
• Convergence rates of Gaussian ODE filters
• Adaptive step-size selection for state-space probabilistic differential equation solvers
• Comment on article by Chkrebtii, Campbell, Calderhead and Girolami

zbMATH Keywords

Gaussian processesMarkov processesfilteringprobabilistic numericsinitial value problemsRunge-Kutta methodsNordsieck methods

Mathematics Subject Classification ID

Computational methods for problems pertaining to statistics (62-08) Multistep, Runge-Kutta and extrapolation methods for ordinary differential equations (65L06) Numerical methods for initial value problems involving ordinary differential equations (65L05)

Cites Work

• Solving Nonstiff Ordinary Differential Equations—The State of the Art
• Stochastic processes and filtering theory
• Mathematical Foundations of Infinite-Dimensional Statistical Models
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Probabilistic numerics and uncertainty in computations
• A Correspondence Between Bayesian Estimation on Stochastic Processes and Smoothing by Splines
• Probability, Frequency and Reasonable Expectation
• Title not available (Why is that?)
• Stochastic differential equations. An introduction with applications.
• Inverse problems: a Bayesian perspective
• Evaluating Derivatives
• Title not available (Why is that?)
• The continuous extension of the B-spline linear multistep methods for BVPs on non-uniform meshes
• Title not available (Why is that?)
• VODE: A Variable-Coefficient ODE Solver
• Title not available (Why is that?)
• B‐Spline Linear Multistep Methods and their Continuous Extensions
• Average case complexity of multivariate integration for smooth functions
• On Numerical Integration of Ordinary Differential Equations
• Comparing Numerical Methods for Ordinary Differential Equations
• A Polyalgorithm for the Numerical Solution of Ordinary Differential Equations
• Analysis of Fixed-Stepsize Methods
• Kalman Filtering
• Chemical instabilities and sustained oscillations
• Order and stepsize control in extrapolation methods
• Title not available (Why is that?)
• Equivalent Forms of Multistep Formulas
• Spline Function Approximations for Solutions of Ordinary Differential Equations
• General linear method: A survey
• Consistency of Nordsieck Methods
• Title not available (Why is that?)
• Bayesian filtering and smoothing
• Title not available (Why is that?)
• Bayesian solution uncertainty quantification for differential equations
• Runge-Kutta Starters for Multistep Methods
• Numerical Solution of Ordinary Differential Equations: Is There Anything Left to Do?
• On Gaussian Optimal Smoothing of Non-Linear State Space Models
• Introduction to Uncertainty Quantification
• Explicit, optimal stability functionals and their application to cyclic discretization methods
• On Testing a Subroutine for the Numerical Integration of Ordinary Differential Equations
• Statistical analysis of differential equations: introducing probability measures on numerical solutions
• The Convergence of Variable-Stepsize, Variable-Formula, Multistep Methods
• Linear Multistep Formulas Based on g-Splines
• Integration Formulas and Schemes Based on g-Splines

Cited In (18)

• GParareal: a time-parallel ODE solver using Gaussian process emulation
• A modern retrospective on probabilistic numerics
• Probabilistic solutions to ordinary differential equations as nonlinear Bayesian filtering: a new perspective
• Randomised one-step time integration methods for deterministic operator differential equations
• A probabilistic finite element method based on random meshes: a posteriori error estimators and Bayesian inverse problems
• Bayesian ODE solvers: the maximum a posteriori estimate
• Bayesian numerical methods for nonlinear partial differential equations
• Estimation of Ordinary Differential Equation Models with Discretization Error Quantification
• Title not available (Why is that?)
• Probabilistic linear solvers: a unifying view
• Modelling the discretization error of initial value problems using the Wishart distribution
• A role for symmetry in the Bayesian solution of differential equations
• Bayesian Probabilistic Numerical Methods
• Convergence rates of Gaussian ODE filters
• A probabilistic representation for the solution of a problem of mathematical physics
• Probabilistic solvers enable a straight-forward exploration of numerical uncertainty in neuroscience models
• Structured learning of rigid‐body dynamics: A survey and unified view from a robotics perspective
• Randomized Quasi-Optimal Local Approximation Spaces in Time

Uses Software

• VODE

This page was built for publication: A probabilistic model for the numerical solution of initial value problems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2329752)