Poisson Random Fields for Dynamic Feature Models

TL;DR

This paper introduces the Wright-Fisher Indian buffet process, a novel Bayesian nonparametric model for capturing time-dependent latent features in data, with applications to dynamic topic modeling of text collections.

Contribution

It develops a new dependent Indian buffet process using Poisson random fields, enabling modeling of evolving features over time in a Bayesian framework.

Findings

Successfully models temporal feature dependencies in text data.

Demonstrates improved topic tracking over time in NIPS papers.

Provides a Markov Chain Monte Carlo inference method for the complex model.

Abstract

We present the Wright-Fisher Indian buffet process (WF-IBP), a probabilistic model for time-dependent data assumed to have been generated by an unknown number of latent features. This model is suitable as a prior in Bayesian nonparametric feature allocation models in which the features underlying the observed data exhibit a dependency structure over time. More specifically, we establish a new framework for generating dependent Indian buffet processes, where the Poisson random field model from population genetics is used as a way of constructing dependent beta processes. Inference in the model is complex, and we describe a sophisticated Markov Chain Monte Carlo algorithm for exact posterior simulation. We apply our construction to develop a nonparametric focused topic model for collections of time-stamped text documents and test it on the full corpus of NIPS papers published from 1987 to 2015.