Output-sensitive adaptive metropolis-hastings for probabilistic programs

David Tolpin; Jan Willem van de Meent; Brooks Paige; Frank Wood

doi:10.1007/978-3-319-23525-7_19

Output-sensitive adaptive metropolis-hastings for probabilistic programs

David Tolpin, Jan Willem van de Meent, Brooks Paige, Frank Wood

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

4 Scopus citations

Abstract

We introduce an adaptive output-sensitive Metropolis-Hastings algorithm for probabilistic models expressed as programs, Adaptive Lightweight Metropolis-Hastings (AdLMH). This algorithm extends Lightweight Metropolis-Hastings (LMH) by adjusting the probabilities of proposing random variables for modification to improve convergence of the program output. We show that AdLMH converges to the correct equilibrium distribution and compare convergence of AdLMH to that of LMH on several test problems to highlight different aspects of the adaptation scheme. We observe consistent improvement in convergence on the test problems.

Original language	English
Title of host publication	Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2015
Editors	Vitor Santos Costa, Carlos Soares, Annalisa Appice, Annalisa Appice, Pedro Pereira Rodrigues, Vitor Santos Costa, Carlos Soares, João Gama, Alípio Jorge, Pedro Pereira Rodrigues, João Gama, Vitor Santos Costa, Alípio Jorge, Annalisa Appice, Pedro Pereira Rodrigues, João Gama, Annalisa Appice, Carlos Soares, Alípio Jorge, João Gama, Pedro Pereira Rodrigues, Vitor Santos Costa, Carlos Soares, Alípio Jorge
Publisher	Springer Verlag
Pages	311-326
Number of pages	16
ISBN (Print)	9783319235240, 9783319235240, 9783319235240, 9783319235240
DOIs	https://doi.org/10.1007/978-3-319-23525-7_19
State	Published - 1 Jan 2015
Externally published	Yes
Event	European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, ECML PKDD 2015 - Porto, Portugal Duration: 7 Sep 2015 → 11 Sep 2015

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	9285
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, ECML PKDD 2015
Country/Territory	Portugal
City	Porto
Period	7/09/15 → 11/09/15

Keywords

Adaptive MCMC
Probabilistic programming

ASJC Scopus subject areas

Theoretical Computer Science
Computer Science (all)

Access to Document

10.1007/978-3-319-23525-7_19

Cite this

Tolpin, D., van de Meent, J. W., Paige, B., & Wood, F. (2015). Output-sensitive adaptive metropolis-hastings for probabilistic programs. In V. S. Costa, C. Soares, A. Appice, A. Appice, P. P. Rodrigues, V. S. Costa, C. Soares, J. Gama, A. Jorge, P. P. Rodrigues, J. Gama, V. S. Costa, A. Jorge, A. Appice, P. P. Rodrigues, J. Gama, A. Appice, C. Soares, A. Jorge, J. Gama, P. P. Rodrigues, V. S. Costa, C. Soares, ... A. Jorge (Eds.), Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2015 (pp. 311-326). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9285). Springer Verlag. https://doi.org/10.1007/978-3-319-23525-7_19

Tolpin, David ; van de Meent, Jan Willem ; Paige, Brooks et al. / Output-sensitive adaptive metropolis-hastings for probabilistic programs. Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2015. editor / Vitor Santos Costa ; Carlos Soares ; Annalisa Appice ; Annalisa Appice ; Pedro Pereira Rodrigues ; Vitor Santos Costa ; Carlos Soares ; João Gama ; Alípio Jorge ; Pedro Pereira Rodrigues ; João Gama ; Vitor Santos Costa ; Alípio Jorge ; Annalisa Appice ; Pedro Pereira Rodrigues ; João Gama ; Annalisa Appice ; Carlos Soares ; Alípio Jorge ; João Gama ; Pedro Pereira Rodrigues ; Vitor Santos Costa ; Carlos Soares ; Alípio Jorge. Springer Verlag, 2015. pp. 311-326 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{79d4d121ffe94f7699e0eec291b370ed,

title = "Output-sensitive adaptive metropolis-hastings for probabilistic programs",

abstract = "We introduce an adaptive output-sensitive Metropolis-Hastings algorithm for probabilistic models expressed as programs, Adaptive Lightweight Metropolis-Hastings (AdLMH). This algorithm extends Lightweight Metropolis-Hastings (LMH) by adjusting the probabilities of proposing random variables for modification to improve convergence of the program output. We show that AdLMH converges to the correct equilibrium distribution and compare convergence of AdLMH to that of LMH on several test problems to highlight different aspects of the adaptation scheme. We observe consistent improvement in convergence on the test problems.",

keywords = "Adaptive MCMC, Probabilistic programming",

author = "David Tolpin and {van de Meent}, {Jan Willem} and Brooks Paige and Frank Wood",

note = "Publisher Copyright: {\textcopyright} Springer International Publishing Switzerland 2015.; European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, ECML PKDD 2015 ; Conference date: 07-09-2015 Through 11-09-2015",

year = "2015",

month = jan,

day = "1",

doi = "10.1007/978-3-319-23525-7_19",

language = "English",

isbn = "9783319235240",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Verlag",

pages = "311--326",

editor = "Costa, {Vitor Santos} and Carlos Soares and Annalisa Appice and Annalisa Appice and Rodrigues, {Pedro Pereira} and Costa, {Vitor Santos} and Carlos Soares and Jo{\~a}o Gama and Al{\'i}pio Jorge and Rodrigues, {Pedro Pereira} and Jo{\~a}o Gama and Costa, {Vitor Santos} and Al{\'i}pio Jorge and Annalisa Appice and Rodrigues, {Pedro Pereira} and Jo{\~a}o Gama and Annalisa Appice and Carlos Soares and Al{\'i}pio Jorge and Jo{\~a}o Gama and Rodrigues, {Pedro Pereira} and Costa, {Vitor Santos} and Carlos Soares and Al{\'i}pio Jorge",

booktitle = "Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2015",

address = "Germany",

}

Tolpin, D, van de Meent, JW, Paige, B & Wood, F 2015, Output-sensitive adaptive metropolis-hastings for probabilistic programs. in VS Costa, C Soares, A Appice, A Appice, PP Rodrigues, VS Costa, C Soares, J Gama, A Jorge, PP Rodrigues, J Gama, VS Costa, A Jorge, A Appice, PP Rodrigues, J Gama, A Appice, C Soares, A Jorge, J Gama, PP Rodrigues, VS Costa, C Soares & A Jorge (eds), Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2015. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 9285, Springer Verlag, pp. 311-326, European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, ECML PKDD 2015, Porto, Portugal, 7/09/15. https://doi.org/10.1007/978-3-319-23525-7_19

Output-sensitive adaptive metropolis-hastings for probabilistic programs. / Tolpin, David; van de Meent, Jan Willem; Paige, Brooks et al.

Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2015. ed. / Vitor Santos Costa; Carlos Soares; Annalisa Appice; Annalisa Appice; Pedro Pereira Rodrigues; Vitor Santos Costa; Carlos Soares; João Gama; Alípio Jorge; Pedro Pereira Rodrigues; João Gama; Vitor Santos Costa; Alípio Jorge; Annalisa Appice; Pedro Pereira Rodrigues; João Gama; Annalisa Appice; Carlos Soares; Alípio Jorge; João Gama; Pedro Pereira Rodrigues; Vitor Santos Costa; Carlos Soares; Alípio Jorge. Springer Verlag, 2015. p. 311-326 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9285).

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

TY - GEN

T1 - Output-sensitive adaptive metropolis-hastings for probabilistic programs

AU - Tolpin, David

AU - van de Meent, Jan Willem

AU - Paige, Brooks

AU - Wood, Frank

N1 - Publisher Copyright: © Springer International Publishing Switzerland 2015.

PY - 2015/1/1

Y1 - 2015/1/1

N2 - We introduce an adaptive output-sensitive Metropolis-Hastings algorithm for probabilistic models expressed as programs, Adaptive Lightweight Metropolis-Hastings (AdLMH). This algorithm extends Lightweight Metropolis-Hastings (LMH) by adjusting the probabilities of proposing random variables for modification to improve convergence of the program output. We show that AdLMH converges to the correct equilibrium distribution and compare convergence of AdLMH to that of LMH on several test problems to highlight different aspects of the adaptation scheme. We observe consistent improvement in convergence on the test problems.

AB - We introduce an adaptive output-sensitive Metropolis-Hastings algorithm for probabilistic models expressed as programs, Adaptive Lightweight Metropolis-Hastings (AdLMH). This algorithm extends Lightweight Metropolis-Hastings (LMH) by adjusting the probabilities of proposing random variables for modification to improve convergence of the program output. We show that AdLMH converges to the correct equilibrium distribution and compare convergence of AdLMH to that of LMH on several test problems to highlight different aspects of the adaptation scheme. We observe consistent improvement in convergence on the test problems.

KW - Adaptive MCMC

KW - Probabilistic programming

UR - http://www.scopus.com/inward/record.url?scp=84959378125&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-23525-7_19

DO - 10.1007/978-3-319-23525-7_19

M3 - Conference contribution

AN - SCOPUS:84959378125

SN - 9783319235240

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 311

EP - 326

BT - Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2015

A2 - Costa, Vitor Santos

A2 - Soares, Carlos

A2 - Appice, Annalisa

A2 - Rodrigues, Pedro Pereira

A2 - Costa, Vitor Santos

A2 - Soares, Carlos

A2 - Gama, João

A2 - Jorge, Alípio

A2 - Rodrigues, Pedro Pereira

A2 - Gama, João

A2 - Costa, Vitor Santos

A2 - Jorge, Alípio

A2 - Appice, Annalisa

A2 - Rodrigues, Pedro Pereira

A2 - Gama, João

A2 - Appice, Annalisa

A2 - Soares, Carlos

A2 - Jorge, Alípio

A2 - Gama, João

A2 - Rodrigues, Pedro Pereira

A2 - Costa, Vitor Santos

A2 - Soares, Carlos

A2 - Jorge, Alípio

PB - Springer Verlag

T2 - European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, ECML PKDD 2015

Y2 - 7 September 2015 through 11 September 2015

ER -

Tolpin D, van de Meent JW, Paige B, Wood F. Output-sensitive adaptive metropolis-hastings for probabilistic programs. In Costa VS, Soares C, Appice A, Appice A, Rodrigues PP, Costa VS, Soares C, Gama J, Jorge A, Rodrigues PP, Gama J, Costa VS, Jorge A, Appice A, Rodrigues PP, Gama J, Appice A, Soares C, Jorge A, Gama J, Rodrigues PP, Costa VS, Soares C, Jorge A, editors, Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2015. Springer Verlag. 2015. p. 311-326. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-319-23525-7_19

Output-sensitive adaptive metropolis-hastings for probabilistic programs

Abstract

Publication series

Conference

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this