[Paper Review] Attractive n-type contact processes

Joseph P. Stover|arXiv (Cornell University)|Jun 29, 2010

Stochastic processes and statistical mechanics11 references3 citations

TL;DR

This paper introduces an interaction map framework to characterize monotonicity (attractiveness) in multitype n-type contact processes, enabling exact simulation via coupling from the past. It proves that a process is attractive if its interaction maps are individually attractive, allowing efficient analysis of complex multitype spatial stochastic models with arbitrary particle type interactions.

ABSTRACT

Interacting particle systems are continuous time Markov processes which are used to construct models in many disciplines. Monotonicity is a property that some interacting particle systems possess. A monotone interacting particle system is called attractive. A benefit of attractiveness is that it simplifies certain calculations, one of which is the ability to use some computational algorithms to sample exactly from the stationary distribution of an ergodic process. Monotonicity is well understood for spin systems such as the contact process. Spin systems only include two particle types however, while in many applied models, it is desirable to include more species of particles. In this paper a general framework of monotonicity will be outlined for a certain class of multitype contact processes.

Motivation & Objective

• To develop a systematic method for assessing monotonicity (attractiveness) in multitype interacting particle systems beyond two-state spin systems.
• To extend the applicability of exact simulation algorithms like coupling from the past (CFTP) to multitype contact processes with three or more particle types.
• To provide a fast, structural criterion—based on interaction maps—for determining whether a multitype contact process is monotone.
• To show that multiple interaction maps, each individually attractive, preserve monotonicity in the overall process.

Proposed method

• The paper introduces an interaction map, a matrix specifying which particle type transitions are allowed and under what conditions, to encode interactions in multitype contact processes.
• It defines a partial order on the state space via particle type ordering, enabling the use of monotonicity criteria for Markov processes.
• Monotonicity is established by verifying that each individual interaction map preserves the partial order under up and down transitions.
• The method uses graphical coupling with independent Poisson point processes for each map’s up and down transitions to maintain stochastic ordering.
• The approach allows decomposition of complex processes into simpler, individually assessable interaction maps.
• The sufficiency condition for monotonicity is proven: if all individual interaction maps are attractive, the full process is monotone.

Experimental results

Research questions

• RQ1 Under what conditions is a multitype contact process with three or more particle types monotone?
• RQ2 Can the attractiveness of a multitype contact process be determined by analyzing its component interaction maps independently?
• RQ3 How can the coupling from the past (CFTP) algorithm be applied to multitype processes that are not spin systems?
• RQ4 What structural properties of interaction maps ensure that the process remains monotone under composition?
• RQ5 Can non-unique or multiple interaction rules between particle types still yield an attractive process?

Key findings

• A multitype contact process is attractive if and only if its interaction map is attractive, providing a necessary and sufficient condition for monotonicity.
• The grass–bushes–trees model is shown to be attractive with no restrictions on rate parameters, using two separate interaction maps.
• The two-stage contact process is monotone with no rate restrictions, confirming its suitability for exact simulation via CFTP.
• Multiple interaction maps can be used to model complex interactions, and if each is individually attractive, the full process remains monotone.
• The interaction map framework allows for rapid assessment of monotonicity without solving complex systems of inequalities.
• The method enables the application of exact sampling algorithms like CFTP to a broad class of multitype spatial stochastic processes previously outside their scope.