Building causal interaction models by recursive unfolding

L.C. van der Gaag; S. Renooij; Alessandro Facchini

Building causal interaction models by recursive unfolding

L.C. van der Gaag
, S. Renooij
, Alessandro Facchini

Dalle Molle Institute for Artificial Intelligence Research
Dalle Molle Institute for Artificial Intelligence Research

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

Abstract

Causal interaction models, such as the well-known noisy-or and leaky noisy-or models, have become quite popular as a means to parameterize conditional probability tables for Bayesian networks. In this paper we focus on the engineering of subnetworks to represent such models and
present a novel technique called recursive unfolding for this purpose. This technique allows inserting, removing and merging cause variables in an interaction model at will, without affecting the
underlying represented information. We detail the technique, with the recursion invariants involved,
and illustrate its practical use for Bayesian-network engineering by means of a small example.

Original language	English
Title of host publication	International Conference on Probabilistic Graphical Models, 23-25 September 2020, Hotel Comwell Rebild Bakker, Skørping, Denmark
Editors	Manfred Jaeger, Thomas Dyhre Nielsen
Publisher	MLResearchPress
Pages	509-520
Number of pages	12
Publication status	Published - 2020
Event	Tenth International Conference on Probabilistic Graphical Models - Hybrid, Aalborg, Denmark Duration: 23 Sept 2020 → 25 Sept 2020 Conference number: 10

Publication series

Name	Proceedings of Machine Learning Research
Publisher	ML Research Press
Volume	138

Conference

Conference	Tenth International Conference on Probabilistic Graphical Models
Abbreviated title	PGM 2020
Country/Territory	Denmark
City	Aalborg
Period	23/09/20 → 25/09/20

Keywords

Bayesian-network engineering
Causal interaction models
Compositionality

Access to Document

vanderGaag20aFinal published version, 494 KBLicence: Taverne

http://proceedings.mlr.press/v138/van-der-gaag20a.html

Cite this

van der Gaag, L. C., Renooij, S., & Facchini, A. (2020). Building causal interaction models by recursive unfolding. In M. Jaeger, & T. D. Nielsen (Eds.), International Conference on Probabilistic Graphical Models, 23-25 September 2020, Hotel Comwell Rebild Bakker, Skørping, Denmark (pp. 509-520). (Proceedings of Machine Learning Research; Vol. 138). MLResearchPress. http://proceedings.mlr.press/v138/van-der-gaag20a.html

van der Gaag, L.C. ; Renooij, S. ; Facchini, Alessandro. / Building causal interaction models by recursive unfolding. International Conference on Probabilistic Graphical Models, 23-25 September 2020, Hotel Comwell Rebild Bakker, Skørping, Denmark. editor / Manfred Jaeger ; Thomas Dyhre Nielsen. MLResearchPress, 2020. pp. 509-520 (Proceedings of Machine Learning Research).

@inproceedings{31e9bc7ab9234392a5ce09ab8f31116c,

title = "Building causal interaction models by recursive unfolding",

abstract = "Causal interaction models, such as the well-known noisy-or and leaky noisy-or models, have become quite popular as a means to parameterize conditional probability tables for Bayesian networks. In this paper we focus on the engineering of subnetworks to represent such models and present a novel technique called recursive unfolding for this purpose. This technique allows inserting, removing and merging cause variables in an interaction model at will, without affecting the underlying represented information. We detail the technique, with the recursion invariants involved, and illustrate its practical use for Bayesian-network engineering by means of a small example. ",

keywords = "Bayesian-network engineering, Causal interaction models, Compositionality",

author = "\{van der Gaag\}, L.C. and S. Renooij and Alessandro Facchini",

year = "2020",

language = "English",

series = "Proceedings of Machine Learning Research",

publisher = "MLResearchPress",

pages = "509--520",

editor = "Manfred Jaeger and Nielsen, \{Thomas Dyhre\}",

booktitle = "International Conference on Probabilistic Graphical Models, 23-25 September 2020, Hotel Comwell Rebild Bakker, Sk{\o}rping, Denmark",

note = "Tenth International Conference on Probabilistic Graphical Models, PGM 2020 ; Conference date: 23-09-2020 Through 25-09-2020",

}

van der Gaag, LC, Renooij, S & Facchini, A 2020, Building causal interaction models by recursive unfolding. in M Jaeger & TD Nielsen (eds), International Conference on Probabilistic Graphical Models, 23-25 September 2020, Hotel Comwell Rebild Bakker, Skørping, Denmark. Proceedings of Machine Learning Research, vol. 138, MLResearchPress, pp. 509-520, Tenth International Conference on Probabilistic Graphical Models, Aalborg, Denmark, 23/09/20. <http://proceedings.mlr.press/v138/van-der-gaag20a.html>

Building causal interaction models by recursive unfolding. / van der Gaag, L.C.; Renooij, S.; Facchini, Alessandro.
International Conference on Probabilistic Graphical Models, 23-25 September 2020, Hotel Comwell Rebild Bakker, Skørping, Denmark. ed. / Manfred Jaeger; Thomas Dyhre Nielsen. MLResearchPress, 2020. p. 509-520 (Proceedings of Machine Learning Research; Vol. 138).

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

TY - GEN

T1 - Building causal interaction models by recursive unfolding

AU - van der Gaag, L.C.

AU - Renooij, S.

AU - Facchini, Alessandro

N1 - Conference code: 10

PY - 2020

Y1 - 2020

N2 - Causal interaction models, such as the well-known noisy-or and leaky noisy-or models, have become quite popular as a means to parameterize conditional probability tables for Bayesian networks. In this paper we focus on the engineering of subnetworks to represent such models and present a novel technique called recursive unfolding for this purpose. This technique allows inserting, removing and merging cause variables in an interaction model at will, without affecting the underlying represented information. We detail the technique, with the recursion invariants involved, and illustrate its practical use for Bayesian-network engineering by means of a small example.

AB - Causal interaction models, such as the well-known noisy-or and leaky noisy-or models, have become quite popular as a means to parameterize conditional probability tables for Bayesian networks. In this paper we focus on the engineering of subnetworks to represent such models and present a novel technique called recursive unfolding for this purpose. This technique allows inserting, removing and merging cause variables in an interaction model at will, without affecting the underlying represented information. We detail the technique, with the recursion invariants involved, and illustrate its practical use for Bayesian-network engineering by means of a small example.

KW - Bayesian-network engineering

KW - Causal interaction models

KW - Compositionality

UR - https://www.semanticscholar.org/paper/Building-Causal-Interaction-Models-by-Recursive-Gaag/701bd3301b3a3ccf2c78674cd4cc058256e83306

M3 - Conference contribution

T3 - Proceedings of Machine Learning Research

SP - 509

EP - 520

BT - International Conference on Probabilistic Graphical Models, 23-25 September 2020, Hotel Comwell Rebild Bakker, Skørping, Denmark

A2 - Jaeger, Manfred

A2 - Nielsen, Thomas Dyhre

PB - MLResearchPress

T2 - Tenth International Conference on Probabilistic Graphical Models

Y2 - 23 September 2020 through 25 September 2020

ER -

Building causal interaction models by recursive unfolding

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this