Efficient bayesian methods for counting processes in partially observable environments

Ferdian Jovan; Jeremy Wyatt; Nick Hawes

Efficient bayesian methods for counting processes in partially observable environments

Ferdian Jovan, Jeremy Wyatt, Nick Hawes

Computer Science

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

51 Downloads (Pure)

Abstract

When sensors that count events are unre- liable, the data sets that result cannot be trusted. We address this common problem by developing practical Bayesian estimators for a partially observable Poisson process (POPP). Unlike Bayesian estimation for a fully observable Poisson process (FOPP) this is non-trivial, since there is no conjugate density for a POPP and the posterior has a number of elements that grow exponentially in the number of observed intervals. We present two tractable approximations, which we combine in a switching filter. This switching filter enables efficient and accurate estimation of the posterior. We perform a detailed empirical analysis, using both simulated and real-
world data.

Original language	English
Title of host publication	Proceedings of the 21st International Conference on Artificial Intelligence and Statistics (AISTATS 2018)
Publisher	Proceedings of Machine Learning Research
Pages	1906-1913
Volume	84
Publication status	E-pub ahead of print - 21 Aug 2018
Event	The 21st International Conference on Artificial Intelligence and Statistics (AISTATS 2018) - Playa Blanca, Lanzarote, Spain Duration: 9 Apr 2018 → 11 Apr 2018

Publication series

Name	Proceedings of Machine Learning Research

Conference

Conference	The 21st International Conference on Artificial Intelligence and Statistics (AISTATS 2018)
Country/Territory	Spain
City	Playa Blanca, Lanzarote
Period	9/04/18 → 11/04/18

Access to Document

Efficient Bayesian
Checked for eligibility: 21/02/2018 Ferdian Jovan, Jeremy Wyatt, Nick Hawes ; Proceedings of the Twenty-First International Conference on Artificial Intelligence and Statistics, PMLR 84:1906-1913, 2018. published online: http://proceedings.mlr.press/v84/jovan18a/jovan18a.pdf Copyright © PMLR 2018. All rights reserved.
Accepted author manuscript, 2.11 MBLicence: None: All rights reserved

http://proceedings.mlr.press/v84/Licence: None: All rights reserved

Cite this

Jovan, F., Wyatt, J., & Hawes, N. (2018). Efficient bayesian methods for counting processes in partially observable environments. In Proceedings of the 21st International Conference on Artificial Intelligence and Statistics (AISTATS 2018) (Vol. 84, pp. 1906-1913). (Proceedings of Machine Learning Research). Proceedings of Machine Learning Research. Advance online publication. http://proceedings.mlr.press/v84/

@inproceedings{ae3f5e6cc3f94e93b741dacb77e27b3c,

title = "Efficient bayesian methods for counting processes in partially observable environments",

abstract = "When sensors that count events are unre- liable, the data sets that result cannot be trusted. We address this common problem by developing practical Bayesian estimators for a partially observable Poisson process (POPP). Unlike Bayesian estimation for a fully observable Poisson process (FOPP) this is non-trivial, since there is no conjugate density for a POPP and the posterior has a number of elements that grow exponentially in the number of observed intervals. We present two tractable approximations, which we combine in a switching filter. This switching filter enables efficient and accurate estimation of the posterior. We perform a detailed empirical analysis, using both simulated and real-world data.",

author = "Ferdian Jovan and Jeremy Wyatt and Nick Hawes",

year = "2018",

month = aug,

day = "21",

language = "English",

volume = "84",

series = "Proceedings of Machine Learning Research",

publisher = "Proceedings of Machine Learning Research",

pages = "1906--1913",

booktitle = "Proceedings of the 21st International Conference on Artificial Intelligence and Statistics (AISTATS 2018)",

note = "The 21st International Conference on Artificial Intelligence and Statistics (AISTATS 2018) ; Conference date: 09-04-2018 Through 11-04-2018",

}

Jovan, F, Wyatt, J & Hawes, N 2018, Efficient bayesian methods for counting processes in partially observable environments. in Proceedings of the 21st International Conference on Artificial Intelligence and Statistics (AISTATS 2018). vol. 84, Proceedings of Machine Learning Research, Proceedings of Machine Learning Research, pp. 1906-1913, The 21st International Conference on Artificial Intelligence and Statistics (AISTATS 2018), Playa Blanca, Lanzarote, Spain, 9/04/18. <http://proceedings.mlr.press/v84/>

Efficient bayesian methods for counting processes in partially observable environments. / Jovan, Ferdian; Wyatt, Jeremy; Hawes, Nick.
Proceedings of the 21st International Conference on Artificial Intelligence and Statistics (AISTATS 2018). Vol. 84 Proceedings of Machine Learning Research, 2018. p. 1906-1913 (Proceedings of Machine Learning Research).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - Efficient bayesian methods for counting processes in partially observable environments

AU - Jovan, Ferdian

AU - Wyatt, Jeremy

AU - Hawes, Nick

PY - 2018/8/21

Y1 - 2018/8/21

N2 - When sensors that count events are unre- liable, the data sets that result cannot be trusted. We address this common problem by developing practical Bayesian estimators for a partially observable Poisson process (POPP). Unlike Bayesian estimation for a fully observable Poisson process (FOPP) this is non-trivial, since there is no conjugate density for a POPP and the posterior has a number of elements that grow exponentially in the number of observed intervals. We present two tractable approximations, which we combine in a switching filter. This switching filter enables efficient and accurate estimation of the posterior. We perform a detailed empirical analysis, using both simulated and real-world data.

AB - When sensors that count events are unre- liable, the data sets that result cannot be trusted. We address this common problem by developing practical Bayesian estimators for a partially observable Poisson process (POPP). Unlike Bayesian estimation for a fully observable Poisson process (FOPP) this is non-trivial, since there is no conjugate density for a POPP and the posterior has a number of elements that grow exponentially in the number of observed intervals. We present two tractable approximations, which we combine in a switching filter. This switching filter enables efficient and accurate estimation of the posterior. We perform a detailed empirical analysis, using both simulated and real-world data.

M3 - Conference contribution

VL - 84

T3 - Proceedings of Machine Learning Research

SP - 1906

EP - 1913

BT - Proceedings of the 21st International Conference on Artificial Intelligence and Statistics (AISTATS 2018)

PB - Proceedings of Machine Learning Research

T2 - The 21st International Conference on Artificial Intelligence and Statistics (AISTATS 2018)

Y2 - 9 April 2018 through 11 April 2018

ER -

Efficient bayesian methods for counting processes in partially observable environments

Abstract

Publication series

Conference

Access to Document

Fingerprint

Cite this