On the relation between filter maps and correction factors in likelihood ratios

The robust form of the likelihood ratio for a signal in the presence of white noise has an additional term in the exponent called the correction factor which corresponds to the trace of the conditional covariance of the signal given the observations. The authors show that this correction term is nothing but the trace of the symmetrized Frechet derivative of the nonlinear filter map and hence the likelihood ratio can be completely represented in terms of the observations and the filter map