TY - JOUR
T1 - Directed Information Graphs
AU - Quinn, Christopher J.
AU - Kiyavash, Negar
AU - Coleman, Todd P.
N1 - Funding Information:
This work was supported in part by the National Science Foundation through the NSF Center for Science of Information under Grant CCF-0939370, Grant CCF-1065352, and Grant SMA-1451221, and in part by the Air Force Office of Scientific Research through MURI under Grant FA9550-10-1-0573 and Grant FA9550-11-1-0016. C. J. Quinn was supported by the U.S. Department of Energy Computational Science Graduate Fellowship under Grant DE-FG02-97ER25308. The authors would like to thank the reviewers for many insightful comments that significantly improved the readability of the manuscript. They would also like to thank one of the reviewers for bringing to their attention an early draft of [39].
Publisher Copyright:
© 1963-2012 IEEE.
PY - 2015/12/1
Y1 - 2015/12/1
N2 - We propose a graphical model for representing networks of stochastic processes, the minimal generative model graph. It is based on reduced factorizations of the joint distribution over time. We show that under appropriate conditions, it is unique and consistent with another type of graphical model, the directed information graph, which is based on a generalization of Granger causality. We demonstrate how directed information quantifies Granger causality in a particular sequential prediction setting. We also develop efficient methods to estimate the topological structure from data that obviate estimating the joint statistics. One algorithm assumes upper bounds on the degrees and uses the minimal dimension statistics necessary. In the event that the upper bounds are not valid, the resulting graph is nonetheless an optimal approximation in terms of Kullback-Leibler (KL) divergence. Another algorithm uses near-minimal dimension statistics when no bounds are known, but the distribution satisfies a certain criterion. Analogous to how structure learning algorithms for undirected graphical models use mutual information estimates, these algorithms use directed information estimates. We characterize the sample-complexity of two plug-in directed information estimators and obtain confidence intervals. For the setting when point estimates are unreliable, we propose an algorithm that uses confidence intervals to identify the best approximation that is robust to estimation error. Last, we demonstrate the effectiveness of the proposed algorithms through the analysis of both synthetic data and real data from the Twitter network. In the latter case, we identify which news sources influence users in the network by merely analyzing tweet times.
AB - We propose a graphical model for representing networks of stochastic processes, the minimal generative model graph. It is based on reduced factorizations of the joint distribution over time. We show that under appropriate conditions, it is unique and consistent with another type of graphical model, the directed information graph, which is based on a generalization of Granger causality. We demonstrate how directed information quantifies Granger causality in a particular sequential prediction setting. We also develop efficient methods to estimate the topological structure from data that obviate estimating the joint statistics. One algorithm assumes upper bounds on the degrees and uses the minimal dimension statistics necessary. In the event that the upper bounds are not valid, the resulting graph is nonetheless an optimal approximation in terms of Kullback-Leibler (KL) divergence. Another algorithm uses near-minimal dimension statistics when no bounds are known, but the distribution satisfies a certain criterion. Analogous to how structure learning algorithms for undirected graphical models use mutual information estimates, these algorithms use directed information estimates. We characterize the sample-complexity of two plug-in directed information estimators and obtain confidence intervals. For the setting when point estimates are unreliable, we propose an algorithm that uses confidence intervals to identify the best approximation that is robust to estimation error. Last, we demonstrate the effectiveness of the proposed algorithms through the analysis of both synthetic data and real data from the Twitter network. In the latter case, we identify which news sources influence users in the network by merely analyzing tweet times.
KW - Causality
KW - directed information
KW - generative models
KW - graphical models
KW - network inference
UR - http://www.scopus.com/inward/record.url?scp=84959449924&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84959449924&partnerID=8YFLogxK
U2 - 10.1109/TIT.2015.2478440
DO - 10.1109/TIT.2015.2478440
M3 - Article
AN - SCOPUS:84959449924
SN - 0018-9448
VL - 61
SP - 6887
EP - 6909
JO - IEEE Transactions on Information Theory
JF - IEEE Transactions on Information Theory
IS - 12
M1 - 7273888
ER -