TY - GEN
T1 - Analyzing community structure in networks
AU - Zhan, Hongyuan
AU - Madduri, Kamesh
N1 - Publisher Copyright:
© 2017 IEEE.
PY - 2017/6/30
Y1 - 2017/6/30
N2 - Real-world networks exhibit significant community structure. Communities are sometimes explicitly known or defined (e.g., virtual groups that one joins in an online social network, departments in an organization), but are often determined using a community detection or a clustering algorithm. Given a weighted network, with edge weights denoting interaction strengths between vertices, and a community membership matrix mapping vertices to overlapping or non-overlapping communities, we present a new unsupervised method for analyzing and ranking these communities, such that the computed nonnegative community weights seek to explain the edge weights. Our method is based on a new factorization of the weighted adjacency matrix. The weighted matrix decomposition we obtain has a simple combinatorial interpretation. We show that the proposed optimization problem reduces to a Nonnegative Least Squares problem, and design a fast algorithm for computing the community weights. We assess this problem formulation on a variety of synthetic and real-world networks, in order to gain insight into its advantages and limitations.
AB - Real-world networks exhibit significant community structure. Communities are sometimes explicitly known or defined (e.g., virtual groups that one joins in an online social network, departments in an organization), but are often determined using a community detection or a clustering algorithm. Given a weighted network, with edge weights denoting interaction strengths between vertices, and a community membership matrix mapping vertices to overlapping or non-overlapping communities, we present a new unsupervised method for analyzing and ranking these communities, such that the computed nonnegative community weights seek to explain the edge weights. Our method is based on a new factorization of the weighted adjacency matrix. The weighted matrix decomposition we obtain has a simple combinatorial interpretation. We show that the proposed optimization problem reduces to a Nonnegative Least Squares problem, and design a fast algorithm for computing the community weights. We assess this problem formulation on a variety of synthetic and real-world networks, in order to gain insight into its advantages and limitations.
UR - http://www.scopus.com/inward/record.url?scp=85028060137&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85028060137&partnerID=8YFLogxK
U2 - 10.1109/IPDPSW.2017.154
DO - 10.1109/IPDPSW.2017.154
M3 - Conference contribution
AN - SCOPUS:85028060137
T3 - Proceedings - 2017 IEEE 31st International Parallel and Distributed Processing Symposium Workshops, IPDPSW 2017
SP - 1540
EP - 1547
BT - Proceedings - 2017 IEEE 31st International Parallel and Distributed Processing Symposium Workshops, IPDPSW 2017
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 31st IEEE International Parallel and Distributed Processing Symposium Workshops, IPDPSW 2017
Y2 - 29 May 2017 through 2 June 2017
ER -