Exact Bayesian inference for level-set Cox processes with piecewise constant intensity function

TL;DR

This paper introduces an exact Bayesian inference method for multidimensional Cox processes with piecewise constant intensity functions, leveraging level-set Gaussian processes and retrospective MCMC techniques for efficient analysis of large datasets.

Contribution

It develops a novel pseudo-marginal MCMC algorithm that performs exact Bayesian inference without discretization, applicable to high-dimensional and spatiotemporal point process models.

Findings

Method achieves exact posterior inference without approximation.

Efficient analysis of large datasets using nearest neighbor Gaussian processes.

Successful application to simulated and real point process data.

Abstract

This paper proposes a new methodology to perform Bayesian inference for a class of multidimensional Cox processes in which the intensity function is piecewise constant. Poisson processes with piecewise constant intensity functions are believed to be suitable to model a variety of point process phenomena and, given its simpler structure, are expected to provide more precise inference when compared to processes with non-parametric and continuously varying intensity functions. The partition of the space domain is flexibly determined by a level-set function of a latent Gaussian process. Despite the intractability of the likelihood function and the infinite dimensionality of the parameter space, inference is performed exactly, in the sense that no space discretization approximation is used and MCMC error is the only source of inaccuracy. That is achieved by using retrospective sampling techniques and devising a pseudo-marginal infinite-dimensional MCMC algorithm that converges to the exact target posterior distribution. Computational efficiency is favored by considering a nearest neighbor Gaussian process, allowing for the analysis of large datasets. An extension to consider spatiotemporal models is also proposed. The efficiency of the proposed methodology is investigated in simulated examples and its applicability is illustrated in the analysis of some real point process datasets.