Estimation for the additive Gaussian channel and Monge-Kantorovitch measure transportation (Q2642038)

The estimation of a signal observed with an additive Gaussian noise is considered in the setting of an abstract Wiener space. More precisely, suppose that the observation is \(Y=X+w\) where \(w\) is a Gaussian variable with values in a separable Fréchet space \(W\), and \(X\) is independent of \(w\) and takes its values in the Cameron-Martin space \(H\) of \(W\). It is proved that the conditional expectation \(E[X| Y]\) can be written as \(\nabla F(Y)\) for some \(H\)-convex function \(F\). The quadratic estimation error is related to the Wasserstein distance between some probability measures.