Implicitly adaptive importance sampling

Abstract Adaptive importance sampling is a class of techniques for finding good proposal distributions sampling. Often the are standard probability whose parameters adapted based on mismatch between current and target distribution. In this work, we present an implicit adaptive method that applies to complicated which not available in closed form. The iteratively matches moments set Monte Carlo draws weighted weights. We apply Bayesian leave-one-out cross-validation show it performs better than many existing parametric methods while being computationally inexpensive.