1: \begin{abstract}

2: %In applications such as finance or security,

3: It is often necessary to make sampling-based statistical inference

4: about many probability distributions in parallel. Given a finite

5: computational resource, this article addresses how to optimally divide

6: sampling effort between the samplers of the different

7: distributions. Formally approaching this decision problem requires

8: both the specification of an error criterion to assess how well each

9: group of samples represent their underlying distribution, and a loss

10: function to combine the errors into an overall performance score. For

11: the first part, a new Monte Carlo divergence error criterion based on

12: Jensen-Shannon divergence is proposed. Using results from information

13: theory, approximations are derived for estimating this criterion

14: for each target based on a single run, enabling adaptive sample size

15: choices to be made during sampling.

16: 

17: %The method is demonstrated on both univariate and variable-dimension

18: %target distributions, including a larger scale example motivated by

19: %network security monitoring.

20: 

21: %Naive strategies for balancing the sample sizes are shown to

22: %be outperformed even by equal sample size allocation, whilst the

23: %proposed method offers robust performance gains.

24: 

25: 

26: \end{abstract}

27: