Automated rejection sampling from product of distributions (Q1775961)

The paper presents approaches for the generation of samples from distributions (possibly multivariate and/or multimodal) with the probability density function (pdf) that are proportional to a product of standard pdfs (for which exact sampling methods are available). The authors demonstrate that in this particular case the rejection sampling procedure can be performed with automated selection of both the comparison function and the upper bound. Moreover, this choice is performed in order to optimise the sampling efficiency among a range of potential solutions.