Speeding up rare event simulations using Kriging models

We consider the Importance Sampling Monte Carlo (ISMC) as a reference probability estimator for estimating very small probabilities in the context of analog circuits design. We propose a surrogate based hybrid ISMC method to accelerate the estimation of probabilities when the budget of simulations is limited. The Kriging model is used as a surrogate of the simulator because it provides the uncertainties around the predictions that are useful to obtain confidence bounds for any model predictions, and consequently for the probability estimators.