Importance sampling based on the kernel density estimator

Abstract

Importance Sampling is an unbiased sampling method used to sample random variables form different densities than originally defined. The importance sampling densities should be constructed to pick up ‘important’ random variables to improve the estimation of a interesting statistics. In this article, we present an importance sampling in which its density function is constructed from the kernel density estimators. This method can generate a sufficient number of samples, and then increase the accuracy of the probability estimate.