English

Quenched large deviations for one dimensional nonlinear filtering

Probability 2014-06-20 v2

Abstract

Consider the standard, one dimensional, nonlinear filtering problem for a diffusion processe Ξt\Xi_t observed in small additive white noise. Denote by q1ϵ()q^\epsilon_1(\cdot) the density of the law of Ξ1\Xi_1 conditioned on σ(Ytϵ:0t1)\sigma(Y_t^\epsilon: 0\leq t\leq 1). We provide "quenched" large deviation estimates for the random family of measures q1ϵ(x)dxq^\epsilon_1(x)dx: there exists a continuous, explicit mapping Jˉ:R2R\bar J : R^2\to R such that for almost all B,VB_\cdot,V_\cdot, Jˉ(,X1)\bar J(\cdot,X_1) is a good rate function and for any measurable GRG\subset R, infxGoJˉ(x,X1)lim infϵlogGq1ϵ(x)dxlim supϵlogGq1ϵ(x)dxinfxGˉJˉ(x,X1).-\inf_{x\in G^o} \bar J(x,X_1) \leq \liminf \epsilon \log \int_G q_1^\epsilon(x) dx \leq \limsup \epsilon \log \int_G q_1^\epsilon(x) dx \leq -\inf_{x\in \bar G} \bar J(x,X_1) .

Keywords

Cite

@article{arxiv.math/0306020,
  title  = {Quenched large deviations for one dimensional nonlinear filtering},
  author = {E. Pardoux and O. Zeitouni},
  journal= {arXiv preprint arXiv:math/0306020},
  year   = {2014}
}

Comments

Corrected version. Same as journal version