Attaining Monotonicity for Bayesian Networks

M.T. Rietbergen; L.C. van der Gaag

Attaining Monotonicity for Bayesian Networks

M.T. Rietbergen, L.C. van der Gaag

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

Abstract

Many real-life Bayesian networks are expected to exhibit commonly known properties of monotonicity, in the sense that higher values for the observable variables should make higher values for the main variable of interest more likely. Yet, violations of these properties may be introduced into a network despite careful engineering efforts. In this paper, we present a method for resolving such violations of monotonicity by varying a single parameter probability. Our method constructs intervals of numerical values to which a parameter can be varied to attain monotonicity without introducing new violations. We argue that our method has a high runtime, yet can be of practical value for specific domains.

Original language	English
Title of host publication	Proceedings 11th European Conference on Symbolic and Quantitative Approaches to Reasoning with Uncertainty (ECSQARU 2011)
Editors	W Liu
Place of Publication	Belfast
Publisher	Springer
Pages	134-145
Number of pages	12
Publication status	Published - 29 Jun 2011

Bibliographical note

European Conference on Symbolic and Quantitative Approaches to Reasoning with Uncertainty

Cite this

@inproceedings{97267e37f8624e31a56276493982c8cd,

title = "Attaining Monotonicity for Bayesian Networks",

abstract = "Many real-life Bayesian networks are expected to exhibit commonly known properties of monotonicity, in the sense that higher values for the observable variables should make higher values for the main variable of interest more likely. Yet, violations of these properties may be introduced into a network despite careful engineering efforts. In this paper, we present a method for resolving such violations of monotonicity by varying a single parameter probability. Our method constructs intervals of numerical values to which a parameter can be varied to attain monotonicity without introducing new violations. We argue that our method has a high runtime, yet can be of practical value for specific domains.",

author = "M.T. Rietbergen and {van der Gaag}, L.C.",

note = "European Conference on Symbolic and Quantitative Approaches to Reasoning with Uncertainty",

year = "2011",

month = jun,

day = "29",

language = "English",

pages = "134--145",

editor = "W Liu",

booktitle = "Proceedings 11th European Conference on Symbolic and Quantitative Approaches to Reasoning with Uncertainty (ECSQARU 2011)",

publisher = "Springer",

}

Attaining Monotonicity for Bayesian Networks. / Rietbergen, M.T.; van der Gaag, L.C.
Proceedings 11th European Conference on Symbolic and Quantitative Approaches to Reasoning with Uncertainty (ECSQARU 2011). ed. / W Liu. Belfast: Springer, 2011. p. 134-145.

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

TY - GEN

T1 - Attaining Monotonicity for Bayesian Networks

AU - Rietbergen, M.T.

AU - van der Gaag, L.C.

N1 - European Conference on Symbolic and Quantitative Approaches to Reasoning with Uncertainty

PY - 2011/6/29

Y1 - 2011/6/29

N2 - Many real-life Bayesian networks are expected to exhibit commonly known properties of monotonicity, in the sense that higher values for the observable variables should make higher values for the main variable of interest more likely. Yet, violations of these properties may be introduced into a network despite careful engineering efforts. In this paper, we present a method for resolving such violations of monotonicity by varying a single parameter probability. Our method constructs intervals of numerical values to which a parameter can be varied to attain monotonicity without introducing new violations. We argue that our method has a high runtime, yet can be of practical value for specific domains.

AB - Many real-life Bayesian networks are expected to exhibit commonly known properties of monotonicity, in the sense that higher values for the observable variables should make higher values for the main variable of interest more likely. Yet, violations of these properties may be introduced into a network despite careful engineering efforts. In this paper, we present a method for resolving such violations of monotonicity by varying a single parameter probability. Our method constructs intervals of numerical values to which a parameter can be varied to attain monotonicity without introducing new violations. We argue that our method has a high runtime, yet can be of practical value for specific domains.

M3 - Conference contribution

SP - 134

EP - 145

BT - Proceedings 11th European Conference on Symbolic and Quantitative Approaches to Reasoning with Uncertainty (ECSQARU 2011)

A2 - Liu, W

PB - Springer

CY - Belfast

ER -

Attaining Monotonicity for Bayesian Networks

Abstract

Bibliographical note

Fingerprint

Cite this