Structure Learning in Causal Cyclic Networks

Sleiman Itani, Mesrob Ohannessian, Karen Sachs, Garry P. Nolan, Munther A. Dahleh
Proceedings of Workshop on Causality: Objectives and Assessment at NIPS 2008, PMLR 6:165-176, 2010.

Abstract

Cyclic graphical models are unnecessary for accurate representation of joint probability distributions, but are often indispensable when a causal representation of variable relationships is desired. For variables with a cyclic causal dependence structure, DAGs are guaranteed not to recover the correct causal structure, and therefore may yield false predictions about the outcomes of perturbations (and even inference.) In this paper, we introduce an approach to generalize Bayesian Network structure learning to structures with cyclic dependence. We introduce a structure learning algorithm, prove its performance given reasonable assumptions, and use simulated data to compare its results to the results of standard Bayesian network structure learning. We then propose a modified, heuristic algorithm with more modest data requirements, and test its performance on a real-life dataset from molecular biology, containing causal, cyclic dependencies.

Cite this Paper

BibTeX
@InProceedings{pmlr-v6-itani10a, title = {Structure Learning in Causal Cyclic Networks}, author = {Itani, Sleiman and Ohannessian, Mesrob and Sachs, Karen and Nolan, Garry P. and Dahleh, Munther A.}, booktitle = {Proceedings of Workshop on Causality: Objectives and Assessment at NIPS 2008}, pages = {165--176}, year = {2010}, editor = {Guyon, Isabelle and Janzing, Dominik and Schölkopf, Bernhard}, volume = {6}, series = {Proceedings of Machine Learning Research}, address = {Whistler, Canada}, month = {12 Dec}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v6/itani10a/itani10a.pdf}, url = {https://proceedings.mlr.press/v6/itani10a.html}, abstract = {Cyclic graphical models are unnecessary for accurate representation of joint probability distributions, but are often indispensable when a causal representation of variable relationships is desired. For variables with a cyclic causal dependence structure, DAGs are guaranteed not to recover the correct causal structure, and therefore may yield false predictions about the outcomes of perturbations (and even inference.) In this paper, we introduce an approach to generalize Bayesian Network structure learning to structures with cyclic dependence. We introduce a structure learning algorithm, prove its performance given reasonable assumptions, and use simulated data to compare its results to the results of standard Bayesian network structure learning. We then propose a modified, heuristic algorithm with more modest data requirements, and test its performance on a real-life dataset from molecular biology, containing causal, cyclic dependencies.} }
Endnote
%0 Conference Paper %T Structure Learning in Causal Cyclic Networks %A Sleiman Itani %A Mesrob Ohannessian %A Karen Sachs %A Garry P. Nolan %A Munther A. Dahleh %B Proceedings of Workshop on Causality: Objectives and Assessment at NIPS 2008 %C Proceedings of Machine Learning Research %D 2010 %E Isabelle Guyon %E Dominik Janzing %E Bernhard Schölkopf %F pmlr-v6-itani10a %I PMLR %P 165--176 %U https://proceedings.mlr.press/v6/itani10a.html %V 6 %X Cyclic graphical models are unnecessary for accurate representation of joint probability distributions, but are often indispensable when a causal representation of variable relationships is desired. For variables with a cyclic causal dependence structure, DAGs are guaranteed not to recover the correct causal structure, and therefore may yield false predictions about the outcomes of perturbations (and even inference.) In this paper, we introduce an approach to generalize Bayesian Network structure learning to structures with cyclic dependence. We introduce a structure learning algorithm, prove its performance given reasonable assumptions, and use simulated data to compare its results to the results of standard Bayesian network structure learning. We then propose a modified, heuristic algorithm with more modest data requirements, and test its performance on a real-life dataset from molecular biology, containing causal, cyclic dependencies.
RIS
TY - CPAPER TI - Structure Learning in Causal Cyclic Networks AU - Sleiman Itani AU - Mesrob Ohannessian AU - Karen Sachs AU - Garry P. Nolan AU - Munther A. Dahleh BT - Proceedings of Workshop on Causality: Objectives and Assessment at NIPS 2008 DA - 2010/02/18 ED - Isabelle Guyon ED - Dominik Janzing ED - Bernhard Schölkopf ID - pmlr-v6-itani10a PB - PMLR DP - Proceedings of Machine Learning Research VL - 6 SP - 165 EP - 176 L1 - http://proceedings.mlr.press/v6/itani10a/itani10a.pdf UR - https://proceedings.mlr.press/v6/itani10a.html AB - Cyclic graphical models are unnecessary for accurate representation of joint probability distributions, but are often indispensable when a causal representation of variable relationships is desired. For variables with a cyclic causal dependence structure, DAGs are guaranteed not to recover the correct causal structure, and therefore may yield false predictions about the outcomes of perturbations (and even inference.) In this paper, we introduce an approach to generalize Bayesian Network structure learning to structures with cyclic dependence. We introduce a structure learning algorithm, prove its performance given reasonable assumptions, and use simulated data to compare its results to the results of standard Bayesian network structure learning. We then propose a modified, heuristic algorithm with more modest data requirements, and test its performance on a real-life dataset from molecular biology, containing causal, cyclic dependencies. ER -
APA
Itani, S., Ohannessian, M., Sachs, K., Nolan, G.P. & Dahleh, M.A.. (2010). Structure Learning in Causal Cyclic Networks. Proceedings of Workshop on Causality: Objectives and Assessment at NIPS 2008, in Proceedings of Machine Learning Research 6:165-176 Available from https://proceedings.mlr.press/v6/itani10a.html.

Related Material