Expectation-Propagation algorithms for linear regression with Poisson noise: application to photon-limited spectral unmixing

Yoann Altmann, Alessandro Perelli, Mike E. Davies

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

This paper discusses Expectation-Propagation (EP) methods for approximate Bayesian inference in the context of linear regression with Poisson noise. We review two main factor graphs used for generalized linear models and discuss how different EP algorithms can be derived. The estimation per- formance based on EP approximations is compared to the per- formance using Monte Carlo sampling from the exact poste- rior distribution. In particular, we observe that using locally independent or isotropic approximate factors enables more robust and scalable algorithms while providing reliable pos- terior means and marginal variances.
LanguageEnglish
Title of host publicationInternational Conference on Acoustics, Speech, and Signal Processing (ICASSP 2019)
Publication statusAccepted/In press - 1 Feb 2019
Event44th IEEE International Conference on Acoustics, Speech, and Signal Processing - Brighton, United Kingdom
Duration: 12 May 201917 May 2019

Conference

Conference44th IEEE International Conference on Acoustics, Speech, and Signal Processing
Abbreviated titleICASSP 2019
CountryUnited Kingdom
CityBrighton
Period12/05/1917/05/19

Fingerprint

Linear regression
Photons
Sampling

Cite this

Altmann, Y., Perelli, A., & Davies, M. E. (Accepted/In press). Expectation-Propagation algorithms for linear regression with Poisson noise: application to photon-limited spectral unmixing. In International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2019)
Altmann, Yoann ; Perelli, Alessandro ; Davies, Mike E. / Expectation-Propagation algorithms for linear regression with Poisson noise: application to photon-limited spectral unmixing. International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2019). 2019.
@inproceedings{37522d0c952e4c149076f4925a1f8b99,
title = "Expectation-Propagation algorithms for linear regression with Poisson noise: application to photon-limited spectral unmixing",
abstract = "This paper discusses Expectation-Propagation (EP) methods for approximate Bayesian inference in the context of linear regression with Poisson noise. We review two main factor graphs used for generalized linear models and discuss how different EP algorithms can be derived. The estimation per- formance based on EP approximations is compared to the per- formance using Monte Carlo sampling from the exact poste- rior distribution. In particular, we observe that using locally independent or isotropic approximate factors enables more robust and scalable algorithms while providing reliable pos- terior means and marginal variances.",
author = "Yoann Altmann and Alessandro Perelli and Davies, {Mike E.}",
year = "2019",
month = "2",
day = "1",
language = "English",
booktitle = "International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2019)",

}

Altmann, Y, Perelli, A & Davies, ME 2019, Expectation-Propagation algorithms for linear regression with Poisson noise: application to photon-limited spectral unmixing. in International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2019). 44th IEEE International Conference on Acoustics, Speech, and Signal Processing, Brighton, United Kingdom, 12/05/19.

Expectation-Propagation algorithms for linear regression with Poisson noise: application to photon-limited spectral unmixing. / Altmann, Yoann; Perelli, Alessandro; Davies, Mike E.

International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2019). 2019.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Expectation-Propagation algorithms for linear regression with Poisson noise: application to photon-limited spectral unmixing

AU - Altmann, Yoann

AU - Perelli, Alessandro

AU - Davies, Mike E.

PY - 2019/2/1

Y1 - 2019/2/1

N2 - This paper discusses Expectation-Propagation (EP) methods for approximate Bayesian inference in the context of linear regression with Poisson noise. We review two main factor graphs used for generalized linear models and discuss how different EP algorithms can be derived. The estimation per- formance based on EP approximations is compared to the per- formance using Monte Carlo sampling from the exact poste- rior distribution. In particular, we observe that using locally independent or isotropic approximate factors enables more robust and scalable algorithms while providing reliable pos- terior means and marginal variances.

AB - This paper discusses Expectation-Propagation (EP) methods for approximate Bayesian inference in the context of linear regression with Poisson noise. We review two main factor graphs used for generalized linear models and discuss how different EP algorithms can be derived. The estimation per- formance based on EP approximations is compared to the per- formance using Monte Carlo sampling from the exact poste- rior distribution. In particular, we observe that using locally independent or isotropic approximate factors enables more robust and scalable algorithms while providing reliable pos- terior means and marginal variances.

M3 - Conference contribution

BT - International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2019)

ER -

Altmann Y, Perelli A, Davies ME. Expectation-Propagation algorithms for linear regression with Poisson noise: application to photon-limited spectral unmixing. In International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2019). 2019