Robust filtering and propagation of uncertainty in hidden Markov models

DOI10.1214/21-EJP633MaRDI QIDQ2042836zbMATH OpenOpenAlexFDO

Authors Andrew L. Allan

Publication date 21 July 2021

Published in Electronic Journal of Probability (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/2005.04982

zbMATH Keywords

filteringhidden Markov modelparameter uncertaintypathwise optimal controlrough equationrough pathsstochastic optimal control

Mathematics Subject Classification ID

Optimality conditions for problems involving randomness (49K45) Signal detection and filtering (aspects of stochastic processes) (60G35) Stochastic ordinary differential equations (aspects of stochastic analysis) (60H10) Markov chains (discrete-time Markov processes on discrete state spaces) (60J10) Rough partial differential equations (60L50) Optimal stochastic control (93E20)

Abstract: We consider the filtering of continuous-time finite-state hidden Markov models, where the rate and observation matrices depend on unknown time-dependent parameters, for which no prior or stochastic model is available. We quantify and analyze how the induced uncertainty may be propagated through time as we collect new observations, and used to simultaneously provide robust estimates of the hidden signal and to learn the unknown parameters, via techniques based on pathwise filtering and new results on the optimal control of rough differential equations.

Review text

The basic problem in filtering of continuous-time finite-state hidden Markov models is to derive an optimal online estimator for an unobserved signal process \(\{X_t\}\) from discrete observations of the path of another process \(\{Y_t\}\) whose dynamics depend on the current state of the signal. Filters are usually sensitive to both uncertainty of the model parameters and to errors in the observed data. One may therefore desire a filter to be robust in both of these distinct senses, namely robust with respect to parameter uncertainty, and continuous with respect to the observation path \(t\to\{Y_t\}\). The author proposes a fully pathwise approach to the problem of filtering and constructs resulting filters which are robust in both of these senses based on new results of the optimal control of rough differential equations introduced by \textit{T. J. Lyons} [Rev. Mat. Iberoam. 14, No. 2, 215--310 (1998; Zbl 0923.34056)] and developed by \textit{J. Diehl} et al. [Appl. Math. Optim. 75, No. 2, 285--315 (2017; Zbl 1373.60099)]. The author's approach is an extension of the one described by himself and \textit{S. N. Cohen} [Ann. Appl. Probab. 30, No. 5, 2274--2310 (2020; Zbl 1471.93265); SIAM J. Control Optim. 57, No. 3, 1646--1671 (2019; Zbl 1420.62407)].

Recommendations

• Finite Horizon Robust State Estimation for Uncertain Finite-Alphabet Hidden Markov Models with Conditional Relative Entropy Constraints
• Robust filtering
• Model robustness of finite state nonlinear filtering over the infinite time horizon
• scientific article; zbMATH DE number 722978
• Finite-dimensional models for hidden Markov chains

Cites work

• A (rough) pathwise approach to a class of non-linear stochastic partial differential equations
• A course on rough paths. With an introduction to regularity structures
• Canonical RDEs and general semimartingales as rough paths
• Controlling rough paths
• Data-driven nonlinear expectations for statistical uncertainty in decisions
• Differential equations driven by rough paths with jumps
• Filtration of a random process in a statistically uncertain linear stochastic differential system
• Fundamentals of stochastic filtering
• scientific article; zbMATH DE number 5871504 (Why is no real title available?)
• Minimax a posteriori estimation of the Markov processes with finite state spaces
• Minimax linear filtering of random sequences with uncertain covariance function
• Minimax linear observers and regulators for stochastic systems with uncertain second-order statistics
• Model robustness of finite state nonlinear filtering over the infinite time horizon
• New results on the innovations problem for non-linear filtering
• On the Bellman equation for some unbounded control problems
• Parameter uncertainty in the Kalman-Bucy filter
• Partial differential equations driven by rough paths
• Pathwise stability of likelihood estimators for diffusions via rough paths
• Pathwise stochastic control with applications to robust filtering
• Robust filtering and prediction for linear systems with uncertain dynamics: A game-theoretic approach
• Robust filtering: correlated noise and multidimensional observation
• Rough path stability of (semi-)linear SPDEs
• Some Applications of Stochastic Differential Equations to Optimal Nonlinear Filtering
• Statistical decision functions which minimize the maximum risk
• Stochastic calculus and applications
• Stochastic control with rough paths
• Stochastic controls. Hamiltonian systems and HJB equations
• The Wonham filter under uncertainty: A game-theoretic approach
• The Wonham Filter With Random Parameters: Rate of Convergence and Error Bounds
• User’s guide to viscosity solutions of second order partial differential equations

Cited in

(8)

• Hidden Markov models: inverse filtering, belief estimation and privacy protection
• Pathwise stochastic control with applications to robust filtering
• scientific article; zbMATH DE number 5979097 (Why is no real title available?)
• Universal Filtering Via Hidden Markov Modeling
• Finite Horizon Robust State Estimation for Uncertain Finite-Alphabet Hidden Markov Models with Conditional Relative Entropy Constraints
• Rough McKean-Vlasov dynamics for robust ensemble Kalman filtering
• Pathwise convergence of the Euler scheme for rough and stochastic differential equations
• Functional differential equations driven by càdlàg rough paths

This page was built for publication: Robust filtering and propagation of uncertainty in hidden Markov models

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2042836)