Scalable Nonparametric Bayesian Inference on Point Processes with Gaussian Processes

TL;DR

This paper introduces a scalable nonparametric Bayesian model using Gaussian Processes for inference on Poisson Point Processes, avoiding gridding and latent thinning, with improved efficiency and accuracy over existing methods.

Contribution

It presents the first nonparametric Bayesian model with linear complexity for Poisson Point Processes using Gaussian Processes, surpassing prior models in speed and accuracy.

Findings

Model has linear complexity and memory requirements.

Faster and more accurate than competing models.

Effectively handles large-scale data.

Abstract

In this paper we propose the first non-parametric Bayesian model using Gaussian Processes to make inference on Poisson Point Processes without resorting to gridding the domain or to introducing latent thinning points. Unlike competing models that scale cubically and have a squared memory requirement in the number of data points, our model has a linear complexity and memory requirement. We propose an MCMC sampler and show that our model is faster, more accurate and generates less correlated samples than competing models on both synthetic and real-life data. Finally, we show that our model easily handles data sizes not considered thus far by alternate approaches.