2008/05/20 by Thomas Bengtsson, Peter Bickel, Bo Li · 2 citations
Mathematics · #math.ST #stat.CO #stat.ME #stat.TH #msc:93E11 #msc:62L12 #msc:86A22 #msc:60G50 #msc:86A32 #msc:86A10
paper · pdf · doi:10.1214/193940307000000518
published as IMS Collections 2008, Vol. 2, 316-334 · Published in at http://dx.doi.org/10.1214/193940307000000518 the IMS Collections (http://www.imstat.org/publications/imscollections.htm) by the Institute of Mathematical Statistics (http://www.imstat.org)
arxiv created 2008/05/20 · arxiv updated 2009/12/01
It has been widely realized that Monte Carlo methods (approximation via a sample ensemble) may fail in large scale systems. This work offers some theoretical insight into this phenomenon in the context of the particle filter. We demonstrate that the maximum of the weights associated with the sample ensemble converges to one as both the sample size and the system dimension tends to infinity. Specifically, under fairly weak assumptions, if the ensemble size grows sub-exponentially in the cube root of the system dimension, the convergence holds for a single update step in state-space models with independent and identically distributed kernels. Further, in an important special case, more refined arguments show (and our simulations suggest) that the convergence to unity occurs unless the ensemble grows super-exponentially in the system dimension. The weight singularity is also established in models with more general multivariate likelihoods, e.g. Gaussian and Cauchy. Although presented in the context of atmospheric data assimilation for numerical weather prediction, our results are generally valid for high-dimensional particle filters.