Robustness of the nonlinear filter (Q1593635): Difference between revisions

The nonlinear filtering model \[ Y_t= \int _0^th(X_s) ds+W_t,\quad 0\leq t\leq T, \] is considered where \(X\) is the signal process, assumed to take values in a complete separable metric space and having r.c.l.l. paths, the observation noise \(W\) is assumed to be an \(\mathbb R^k\)-valued Brownian motion, \(h\) is a measurable function and \(Y\) is the observation process. The process \(X\) is not assumed to be Markov. Under some integrability hypothesis the filter is shown to depend continuously on the law of the signal. The analysis is based on expressing the nonlinear filter as a Wiener functional via the Kallianpur-Striebel Bayes formula.