1: \begin{abstract}

2: This paper deals with the estimation of rare event probabilities using importance sampling (IS), where an {\it optimal} proposal distribution is computed with the cross-entropy (CE) method. Although, IS optimized with the CE method leads to an efficient reduction of the estimator  variance, this approach  remains  unaffordable for problems where the repeated evaluation of the score function represents a too intensive computational effort.   This is often the case  for score functions related to the solution of a partial differential equation (PDE)  with random inputs. 

3: This work proposes to alleviate  computation  by the parsimonious use of a hierarchy of score function approximations  in the CE optimization process. The score function approximation is obtained by selecting the surrogate of lowest dimensionality, whose accuracy guarantees to pass the  current CE optimization stage.  The selection  relies on certified upper bounds on the  error norm. An asymptotic analysis provides some theoretical guarantees on the efficiency and convergence of the proposed algorithm. Numerical results demonstrate the gain brought by the  method in the context of pollution alerts and a system modeled by a PDE.

4: 

5: 

6: \end{abstract}

7: