1: \begin{abstract}

2:   Adaptive importance sampling is a powerful tool to sample from complicated

3:   target densities, but its success depends sensitively on the initial proposal

4:   density. An algorithm is presented to automatically perform the initialization

5:   using Markov chains and hierarchical clustering. The performance is checked on

6:   challenging multimodal examples in up to 20 dimensions and compared to results

7:   from nested sampling. Our approach yields a proposal that leads to rapid

8:   convergence and accurate estimation of overall normalization and marginal

9:   distributions.

10: 

11: \end{abstract}

12: