Change-point detection for Piecewise Deterministic Markov Processes

TL;DR

This paper develops a numerical method for detecting change points in Piecewise Deterministic Markov Processes observed with noise, by formulating it as an optimal stopping problem and using quantization for approximation.

Contribution

It introduces a discretization-based approach to detect change points in PDMPs, including error bounds and numerical validation of the method.

Findings

Effective change-point detection in noisy PDMP observations

Quantization-based approximation provides accurate value function estimates

Numerical simulations demonstrate practical performance of the proposed policy

Abstract

We consider a change-point detection problem for a simple class of Piecewise Deterministic Markov Processes (PDMPs). A continuous-time PDMP is observed in discrete time and through noise, and the aim is to propose a numerical method to accurately detect both the date of the change of dynamics and the new regime after the change. To do so, we state the problem as an optimal stopping problem for a partially observed discrete-time Markov decision process taking values in a continuous state space and provide a discretization of the state space based on quantization to approximate the value function and build a tractable stopping policy. We provide error bounds for the approximation of the value function and numerical simulations to assess the performance of our candidate policy.