Introduction

Perfect simulation algorithms draw random variates from a distribution using a random number of steps. The term perfect simulation can be used either as an adjective for a specific algorithm with these properties, or for a general protocol for creating algorithms with these properties. Perfect simulation algorithms have been developed for Markov random fields, permutation problems, spatial point processes, stochastic differential equations, and many other applications. Until the advent of perfect simulation, Markov chain Monte Carlo was often used to draw approximate samples from these distributions.