1: \begin{abstract}

2:   Markov chain Monte Carlo methods are a powerful and commonly used

3:   family of numerical methods for sampling from complex probability

4:   distributions. As applications of these methods increase in size and

5:   complexity, the need for efficient methods which can exploit the

6:   parallel architectures which are prevalent in high performance

7:   computing increases. In this paper, we aim to develop a framework

8:   for scalable parallel MCMC algorithms. At each iteration, an

9:   importance sampling proposal distribution is formed using the

10:   current states of all of the chains within an ensemble. Once

11:   weighted samples have been produced from this, a state-of-the-art

12:   resampling method is then used to create an evenly weighted sample

13:   ready for the next iteration. We demonstrate that this parallel

14:   adaptive importance sampling (PAIS) method outperforms naive

15:   parallelisation of serial MCMC methods using the same number of

16:   ensemble members, for low dimensional problems, and in fact shows

17:   better than linear improvements in convergence rates with respect to

18:   the number of ensemble members. We also introduce a new resampling

19:   strategy, approximate multinomial resampling (AMR), which while not

20:   as accurate as other schemes is substantially less costly for large

21:   ensemble sizes, which can then be used in conjunction with PAIS for

22:   complex problems.

23: \end{abstract}