A general context-tree-based approach to perfect simulation for chains of infinite order

Authors:
  • SGSandro Gallo
    Sandro Gallo
  • NGNancy L. Garcia
    Nancy L. Garcia

Abstract
We introduce a general framework for perfect simulation of chains of infinite order on countable alphabet. It is based on the fact that the class of locally continuous probabilistic kernels corresponds to the class of kernels that can be decomposed as a mixture of probabilistic context trees. Our method not only works to perfectly simulate discontinuous chains but also allows to perfectly simulate continuous chains faster than the existing algorithms of the literature.