Generating Random Elements of Finite Distributive Lattices

James Propp

Abstract


This survey article describes a method for choosing uniformly at random from any finite set whose objects can be viewed as constituting a distributive lattice. The method is based on ideas of the author and David Wilson for using "coupling from the past" to remove initialization bias from Monte Carlo randomization. The article describes several applications to specific kinds of combinatorial objects such as tilings, constrained lattice paths, an alternating sign matrices.


Full Text: PDF COMMENT