Adaptive online variance estimation in particle filters: the ALVar estimator

TL;DR

This paper introduces the ALVar estimator, an adaptive online method for real-time variance estimation in particle filters that is consistent, parameter-free, and computationally efficient.

Contribution

The paper presents the ALVar estimator, a novel adaptive online variance estimation technique for particle filters that requires no parameter calibration and operates efficiently in real time.

Findings

ALVar estimator is consistent for true asymptotic variance.

It operates with constant average computational complexity per iteration.

Numerical results show ALVar's superiority over existing methods.

Abstract

We present a new approach-the ALVar estimator-to estimation of asymptotic variance in sequential Monte Carlo methods, or, particle filters. The method, which adjusts adaptively the lag of the estimator proposed in [Olsson, J. and Douc, R. (2019). Numerically stable online estimation of variance in particle filters. Bernoulli, 25(2), pp. 1504-1535] applies to very general distribution flows and particle filters, including auxiliary particle filters with adaptive resampling. The algorithm operates entirely online, in the sense that it is able to monitor the variance of the particle filter in real time and with, on the average, constant computational complexity and memory requirements per iteration. Crucially, it does not require the calibration of any algorithmic parameter. Estimating the variance only on the basis of the genealogy of the propagated particle cloud, without additional simulations, the routine requires only minor code additions to the underlying particle algorithm. Finally, we prove that the ALVar estimator is consistent for the true asymptotic variance as the number of particles tends to infinity and illustrate numerically its superiority to existing approaches.