The combinatorial structure of beta negative binomial processes

DOI10.3150/15-BEJ729MaRDI QIDQ726741zbMATH OpenFDO

Authors Creighton Heaukulani, Daniel M. Roy

Publication date 14 July 2016

Published in Bernoulli (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1401.0062, https://projecteuclid.org/euclid.bj/1462297682

zbMATH Keywords

Bayesian nonparametricsbeta negative binomial processesIndian buffet processlatent feature modelsMarkov chain Monte Carlo algorithmmultisets

Mathematics Subject Classification ID

Exchangeability for stochastic processes (60G09) Point processes (e.g., Poisson, Cox, Hawkes processes) (60G55) Random measures (60G57) Stochastic processes (60G99) Markov chains (discrete-time Markov processes on discrete state spaces) (60J10) Computational methods in Markov chains (60J22) Monte Carlo methods (65C05) Numerical analysis or methods applied to Markov chains (65C40)

Abstract: We characterize the combinatorial structure of conditionally-i.i.d. sequences of negative binomial processes with a common beta process base measure. In Bayesian nonparametric applications, such processes have served as models for latent multisets of features underlying data. Analogously, random subsets arise from conditionally-i.i.d. sequences of Bernoulli processes with a common beta process base measure, in which case the combinatorial structure is described by the Indian buffet process. Our results give a count analogue of the Indian buffet process, which we call a negative binomial Indian buffet process. As an intermediate step toward this goal, we provide a construction for the beta negative binomial process that avoids a representation of the underlying beta process base measure. We describe the key Markov kernels needed to use a NB-IBP representation in a Markov Chain Monte Carlo algorithm targeting a posterior distribution.

Review text

This paper treats a characterization for the combinatorial structure of conditionally i.i.d sequences of negative binomial processes \(X_j\) \(\sim\) \(\mathrm{NBP}(r, B_0)\) with a common beta process base measure \(B_0\) on \(( \Omega, {\mathcal A})\), where \(r > 0\) is a parameter. In a Bayesian non-parametric applicatory framework, such processes have served as models for latent multisets of features underlying certain data. Analogously, random subsets arise from conditionally i.i.d. sequences of Bernoulli processes with a common beta process base measure. In this case, the combinatorial structure is described by the Indian buffet process (IBP). The principal feature of the proof consists in the description of the key Markov kernels needed to use a negative binomial Indian buffet process (NB-IBP) representation in a Markov chain Monte Carlo algorithm targeting a posterior distribution. For other related works, see, e.g., [\textit{T. Broderick} et al., ``Combinatorial clustering and the beta-negative binomial process, IEEE Trans. Pattern Anal. Mach. Intell. 37, No. 2, 290--306, (2014; \url{doi:10.1109/TPAMI.2014.2318721}); \textit{N. L. Hjort}, Ann. Stat. 18, No. 3, 1259--1294 (1990; Zbl 0711.62033); \textit{G. Gregoire}, Stochastic Processes Appl. 16, 179--188 (1984; Zbl 0522.60048)].

Recommendations

• Cluster and feature modeling from combinatorial stochastic processes
• Beta processes, stick-breaking and power laws
• Nonparametric Bayesian negative binomial factor analysis
• The Indian buffet process: an introduction and review
• Gibbs-type Indian buffet processes

Cites work

• Bayesian nonparametric statistical inference for Poisson point processes
• Cluster and feature modeling from combinatorial stochastic processes
• Completely random measures
• Distributional properties of the negative binomial Lévy process
• Feature allocations, probability functions, and paintboxes
• Generalized hypergeometric, digamma and trigamma distributions
• scientific article; zbMATH DE number 1713116 (Why is no real title available?)
• scientific article; zbMATH DE number 6114082 (Why is no real title available?)
• Negative binomial distributions for point processes
• Negative binomial processes
• Nonparametric Bayes estimators based on beta processes in models for life history data
• Nonparametric Bayesian estimators for counting processes
• Poisson/gamma random field models for spatial statistics
• Slice sampling. (With discussions and rejoinder)

Cited in

(3)

• Nonparametric Bayesian negative binomial factor analysis
• Beta processes, stick-breaking and power laws
• Transform-scaled process priors for trait allocations in Bayesian nonparametrics

This page was built for publication: The combinatorial structure of beta negative binomial processes

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q726741)