Robustness in sum-product networks with continuous and categorical data

Rob C. de Wit; Cassio P. de Campos; Diarmaid Conaty; Jesús Martínez del Rincon

Robustness in sum-product networks with continuous and categorical data

Rob C. de Wit, Cassio P. de Campos, Diarmaid Conaty, Jesús Martínez del Rincon

Uncertainty in Artificial Intelligence

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

1 Citation (Scopus)

Abstract

Sum-product networks are a popular family of probabilistic graphical models for which marginal inference can be performed in polynomial time. After learning sum-product networks from scarce data, small variations of parameters could lead to different conclusions. We adapt the robustness measure created for categorical credal sum-product networks to domains with both continuous and categorical variables. We apply this approach to a real-world dataset of online purchases where the goal is to identify fraudulent cases. We empirically show that such credal models can better discriminate between easy and hard instances than simply using the probability of the most probable class.

Original language	English
Title of host publication	Proceedings of the Eleventh International Symposium on Imprecise Probabilities: Theories and Applications
Editors	Jasper De Bock, Cassio P. de Campos, Gert de Cooman, Erik Quaeghebeur, Gregory Wheeler
Publisher	PMLR
Pages	156-158
Number of pages	3
Publication status	Published - 2019
Event	11th International Symposium on Imprecise Probability: Theories and Applications, ISIPTA 2019 - Ghent, Belgium Duration: 3 Jul 2019 → 6 Jul 2019 Conference number: 11

Publication series

Name	Proceedings of Machine Learning Research
Publisher	PMLR
Volume	103

Conference

Conference	11th International Symposium on Imprecise Probability: Theories and Applications, ISIPTA 2019
Abbreviated title	ISIPTA 2019
Country/Territory	Belgium
City	Ghent
Period	3/07/19 → 6/07/19

Access to Document

http://proceedings.mlr.press/v103/de-wit19a.html

Cite this

de Wit, R. C., de Campos, C. P., Conaty, D., & Martínez del Rincon, J. (2019). Robustness in sum-product networks with continuous and categorical data. In J. De Bock, C. P. de Campos, G. de Cooman, E. Quaeghebeur, & G. Wheeler (Eds.), Proceedings of the Eleventh International Symposium on Imprecise Probabilities: Theories and Applications (pp. 156-158). (Proceedings of Machine Learning Research; Vol. 103). PMLR. http://proceedings.mlr.press/v103/de-wit19a.html

de Wit, Rob C. ; de Campos, Cassio P. ; Conaty, Diarmaid et al. / Robustness in sum-product networks with continuous and categorical data. Proceedings of the Eleventh International Symposium on Imprecise Probabilities: Theories and Applications. editor / Jasper De Bock ; Cassio P. de Campos ; Gert de Cooman ; Erik Quaeghebeur ; Gregory Wheeler. PMLR, 2019. pp. 156-158 (Proceedings of Machine Learning Research).

@inproceedings{a3c30019810b412bbf89492272571623,

title = "Robustness in sum-product networks with continuous and categorical data",

abstract = "Sum-product networks are a popular family of probabilistic graphical models for which marginal inference can be performed in polynomial time. After learning sum-product networks from scarce data, small variations of parameters could lead to different conclusions. We adapt the robustness measure created for categorical credal sum-product networks to domains with both continuous and categorical variables. We apply this approach to a real-world dataset of online purchases where the goal is to identify fraudulent cases. We empirically show that such credal models can better discriminate between easy and hard instances than simply using the probability of the most probable class.",

author = "{de Wit}, {Rob C.} and {de Campos}, {Cassio P.} and Diarmaid Conaty and {Mart{\'i}nez del Rincon}, Jes{\'u}s",

year = "2019",

language = "English",

series = "Proceedings of Machine Learning Research",

publisher = "PMLR",

pages = "156--158",

editor = "{De Bock}, Jasper and {de Campos}, {Cassio P.} and {de Cooman}, Gert and Erik Quaeghebeur and Gregory Wheeler",

booktitle = "Proceedings of the Eleventh International Symposium on Imprecise Probabilities: Theories and Applications",

note = "11th International Symposium on Imprecise Probability: Theories and Applications, ISIPTA 2019, ISIPTA 2019 ; Conference date: 03-07-2019 Through 06-07-2019",

}

de Wit, RC, de Campos, CP, Conaty, D & Martínez del Rincon, J 2019, Robustness in sum-product networks with continuous and categorical data. in J De Bock, CP de Campos, G de Cooman, E Quaeghebeur & G Wheeler (eds), Proceedings of the Eleventh International Symposium on Imprecise Probabilities: Theories and Applications. Proceedings of Machine Learning Research, vol. 103, PMLR, pp. 156-158, 11th International Symposium on Imprecise Probability: Theories and Applications, ISIPTA 2019, Ghent, Belgium, 3/07/19. <http://proceedings.mlr.press/v103/de-wit19a.html>

Robustness in sum-product networks with continuous and categorical data. / de Wit, Rob C.; de Campos, Cassio P.; Conaty, Diarmaid et al.
Proceedings of the Eleventh International Symposium on Imprecise Probabilities: Theories and Applications. ed. / Jasper De Bock; Cassio P. de Campos; Gert de Cooman; Erik Quaeghebeur; Gregory Wheeler. PMLR, 2019. p. 156-158 (Proceedings of Machine Learning Research; Vol. 103).

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

TY - GEN

T1 - Robustness in sum-product networks with continuous and categorical data

AU - de Wit, Rob C.

AU - de Campos, Cassio P.

AU - Conaty, Diarmaid

AU - Martínez del Rincon, Jesús

N1 - Conference code: 11

PY - 2019

Y1 - 2019

N2 - Sum-product networks are a popular family of probabilistic graphical models for which marginal inference can be performed in polynomial time. After learning sum-product networks from scarce data, small variations of parameters could lead to different conclusions. We adapt the robustness measure created for categorical credal sum-product networks to domains with both continuous and categorical variables. We apply this approach to a real-world dataset of online purchases where the goal is to identify fraudulent cases. We empirically show that such credal models can better discriminate between easy and hard instances than simply using the probability of the most probable class.

AB - Sum-product networks are a popular family of probabilistic graphical models for which marginal inference can be performed in polynomial time. After learning sum-product networks from scarce data, small variations of parameters could lead to different conclusions. We adapt the robustness measure created for categorical credal sum-product networks to domains with both continuous and categorical variables. We apply this approach to a real-world dataset of online purchases where the goal is to identify fraudulent cases. We empirically show that such credal models can better discriminate between easy and hard instances than simply using the probability of the most probable class.

M3 - Conference contribution

T3 - Proceedings of Machine Learning Research

SP - 156

EP - 158

BT - Proceedings of the Eleventh International Symposium on Imprecise Probabilities: Theories and Applications

A2 - De Bock, Jasper

A2 - de Campos, Cassio P.

A2 - de Cooman, Gert

A2 - Quaeghebeur, Erik

A2 - Wheeler, Gregory

PB - PMLR

T2 - 11th International Symposium on Imprecise Probability: Theories and Applications, ISIPTA 2019

Y2 - 3 July 2019 through 6 July 2019

ER -

de Wit RC, de Campos CP, Conaty D, Martínez del Rincon J. Robustness in sum-product networks with continuous and categorical data. In De Bock J, de Campos CP, de Cooman G, Quaeghebeur E, Wheeler G, editors, Proceedings of the Eleventh International Symposium on Imprecise Probabilities: Theories and Applications. PMLR. 2019. p. 156-158. (Proceedings of Machine Learning Research).

Robustness in sum-product networks with continuous and categorical data

Abstract

Publication series

Conference

Access to Document

Fingerprint

Cite this