Granger-causality meets causal inference in graphical models: Learning networks via non-invasive observations

Mihaela Dimovska, Donatello Materassi

Research output: Chapter in Book/Report/Conference proceedingConference contribution

7 Scopus citations

Abstract

Algorithms developed in the area of graphical models have proven to be useful tools for the analysis and reconstruction of networks of dynamic systems with directed acyclic structure. However, such techniques cannot typically provide a consistent reconstruction for networks of dynamic systems in presence of feedback mechanisms. On the other hand, reconstruction techniques based on causal estimators and one-step-ahead predictors, such as Granger causality, are capable of reconstructing loopy networks, but only if all the operators defining the input/output structure of the network are strictly causal. In this work, we develop a novel reconstruction method for network topologies that combines the advantages of graphical models and causal estimator techniques unifying both approaches under a single framework. The fundamental result is a generalization of Granger causality which provides a consistent reconstruction of the topology of a network of linear dynamic systems under the milder assumption that every loop contains at least one strictly causal transfer function.

Original languageEnglish (US)
Title of host publication2017 IEEE 56th Annual Conference on Decision and Control, CDC 2017
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages5268-5273
Number of pages6
ISBN (Electronic)9781509028733
DOIs
StatePublished - Jun 28 2017
Externally publishedYes
Event56th IEEE Annual Conference on Decision and Control, CDC 2017 - Melbourne, Australia
Duration: Dec 12 2017Dec 15 2017

Publication series

Name2017 IEEE 56th Annual Conference on Decision and Control, CDC 2017
Volume2018-January

Other

Other56th IEEE Annual Conference on Decision and Control, CDC 2017
Country/TerritoryAustralia
CityMelbourne
Period12/12/1712/15/17

Bibliographical note

Funding Information:
*This work is partially supported by NSF (CISE/CAREER:1553504).

Publisher Copyright:
© 2017 IEEE.

Fingerprint

Dive into the research topics of 'Granger-causality meets causal inference in graphical models: Learning networks via non-invasive observations'. Together they form a unique fingerprint.

Cite this