Abstract:Abstract This article shows that increasing the observation variance at small scales can reduce the ensemble size required to avoid collapse in particle filtering of spatially extended dynamics and improve the resulting uncertainty quantification at large scales. Particle filter weights depend on how well ensemble members agree with observations, and collapse occurs when a few ensemble members receive most of the weight. Collapse causes catastrophic variance underestimation. Increasing small-scale variance in the observation error model reduces the incidence of collapse by de-emphasizing small-scale differences between the ensemble members and the observations. Doing so smooths the posterior mean, though it does not smooth the individual ensemble members. Two options for implementing the proposed observation error model are described. Taking a discretized elliptic differential operator as an observation error covariance matrix provides the desired property of a spectrum that grows in the approach to small scales. This choice also introduces structure exploitable by scalable computation techniques, including multigrid solvers and multiresolution approximations to the corresponding integral operator. Alternatively the observations can be smoothed and then assimilated under the assumption of independent errors, which is equivalent to assuming large errors at small scales. The method is demonstrated on a linear stochastic partial differential equation, where it significantly reduces the occurrence of particle filter collapse while maintaining accuracy. It also improves continuous ranked probability scores by as much as 25%, indicating that the weighted ensemble more accurately represents the true distribution. The method is compatible with other techniques for improving the performance of particle filters.

Curse-of-dimensionality revisited: Collapse of importance sampling in very large scale systems

Curse-of-dimensionality revisited: Collapse of the particle filter in very large scale systems

Importance Sampling for the Empirical Measure of Weakly Interacting Diffusions

On importance sampling and independent Metropolis-Hastings with an unbounded weight function

Achieving High Convergence Rates by Quasi-Monte Carlo and Importance Sampling for Unbounded Integrands

Particle Efficient Importance Sampling

Efficient large deviation estimation based on importance sampling

Dynamical computation of the density of states and Bayes factors using nonequilibrium importance sampling

Iterative Importance Sampling Algorithms for Parameter Estimation

Quasi-Monte Carlo for unbounded integrands with importance sampling

Non-Asymptotic Analysis of Ensemble Kalman Updates: Effective Dimension and Localization

Scalability of Metropolis-within-Gibbs schemes for high-dimensional Bayesian models

Improving Particle Filter Performance by Smoothing Observations

Randomly Shifted Lattice Rules with Importance Sampling and Applications

Breaking the Curse for Uniform Approximation in Hilbert Spaces via Monte Carlo Methods

Cross-entropy-based importance sampling with failure-informed dimension reduction for rare event simulation

Importance sampling for counting statistics in one-dimensional systems

Sampling normalizing constants in high dimensions using inhomogeneous diffusions

Efficient High-Dimensional Importance Sampling

Using Autodiff to Estimate Posterior Moments, Marginals and Samples