Improvements to exact Boltzmann sampling using probabilistic divide-and-conquer and the recursive method
14 jul 2017
Acerca de este artículo
Publicado en línea: 14 jul 2017
Páginas: 22 - 45
Recibido: 30 sept 2016
DOI: https://doi.org/10.1515/puma-2015-0020
Palabras clave
© 2017
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.
We demonstrate an approach for exact sampling of certain discrete combinatorial distributions, which is a hybrid of exact Boltzmann sampling and the recursive method, using probabilistic divide-and-conquer (PDC). The approach specializes to exact Boltzmann sampling in the trivial setting, and specializes to PDC deterministic second half in the first non-trivial application. A large class of examples is given for which this method broadly applies, and several examples are worked out explicitly.