Particle filters rely on resampling to avoid weight degeneracy, but in high dimensions this operation distorts the spectrum of the ensemble covariance. The distortion is not arbitrary; it can be computed from the importance weights, which opens the door to a principled correction.

The new paper, "Regularized free deconvolution for spectral repair of resampling distortion in high-dimensional particle filters," addresses this problem directly. As the title suggests, the proposed approach uses regularized free deconvolution to repair the spectrum after resampling.

The abstract is truncated, so full algorithmic details and experimental results are not available from the source. Still, the paper appears to offer a systematic way to undo a known distortion, rather than relying on ad hoc fixes.