|
Abstract:
|
We propose a recursive algorithm to estimate a finite set of conditional distributions. The procedure is fully nonparametric and has a Bayesian interpretation. Indeed, the recursive process follows a certain Bayesian update. We prove weak convergence of the sequence of distributions, using a fixed point argument. This asymptotic result is new in the context of recursive algorithms. We demonstrate numerical accuracy via simulations. The estimate is very fast, it is sequential and requires limited computing power; being also parallelizable.
|
