On triangular versus edge representations - Towards scalable modeling of networks

Qirong Ho, Junming Yin, Eric P. Xing

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

11 Citations (Scopus)

Abstract

In this paper, we argue for representing networks as a bag of triangular motifs, particularly for important network problems that current model-based approaches handle poorly due to computational bottlenecks incurred by using edge representations. Such approaches require both 1-edges and 0-edges (missing edges) to be provided as input, and as a consequence, approximate inference algorithms for these models usually require Ω(N2) time per iteration, precluding their application to larger real-world networks. In contrast, triangular modeling requires less computation, while providing equivalent or better inference quality. A triangular motif is a vertex triple containing 2 or 3 edges, and the number of such motifs is Θ(Σ i Di2) (where Di is the degree of vertex i), which is much smaller than N2 for low-maximum-degree networks. Using this representation, we develop a novel mixed-membership network model and approximate inference algorithm suitable for large networks with low max-degree. For networks with high maximum degree, the triangular motifs can be naturally subsampled in a node-centric fashion, allowing for much faster inference at a small cost in accuracy. Empirically, we demonstrate that our approach, when compared to that of an edge-based model, has faster runtime and improved accuracy for mixed-membership community detection. We conclude with a large-scale demonstration on an N ≈ 280, 000-node network, which is infeasible for network models with (N2) inference cost.

Original languageEnglish (US)
Title of host publicationAdvances in Neural Information Processing Systems
Pages2132-2140
Number of pages9
Volume3
StatePublished - 2012
Externally publishedYes
Event26th Annual Conference on Neural Information Processing Systems 2012, NIPS 2012 - Lake Tahoe, NV, United States
Duration: Dec 3 2012Dec 6 2012

Other

Other26th Annual Conference on Neural Information Processing Systems 2012, NIPS 2012
CountryUnited States
CityLake Tahoe, NV
Period12/3/1212/6/12

Fingerprint

Costs
Demonstrations

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Information Systems
  • Signal Processing

Cite this

Ho, Q., Yin, J., & Xing, E. P. (2012). On triangular versus edge representations - Towards scalable modeling of networks. In Advances in Neural Information Processing Systems (Vol. 3, pp. 2132-2140)

On triangular versus edge representations - Towards scalable modeling of networks. / Ho, Qirong; Yin, Junming; Xing, Eric P.

Advances in Neural Information Processing Systems. Vol. 3 2012. p. 2132-2140.

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

Ho, Q, Yin, J & Xing, EP 2012, On triangular versus edge representations - Towards scalable modeling of networks. in Advances in Neural Information Processing Systems. vol. 3, pp. 2132-2140, 26th Annual Conference on Neural Information Processing Systems 2012, NIPS 2012, Lake Tahoe, NV, United States, 12/3/12.
Ho Q, Yin J, Xing EP. On triangular versus edge representations - Towards scalable modeling of networks. In Advances in Neural Information Processing Systems. Vol. 3. 2012. p. 2132-2140
Ho, Qirong ; Yin, Junming ; Xing, Eric P. / On triangular versus edge representations - Towards scalable modeling of networks. Advances in Neural Information Processing Systems. Vol. 3 2012. pp. 2132-2140
@inproceedings{10b5ae87a8ba4c8a95e47112c1189918,
title = "On triangular versus edge representations - Towards scalable modeling of networks",
abstract = "In this paper, we argue for representing networks as a bag of triangular motifs, particularly for important network problems that current model-based approaches handle poorly due to computational bottlenecks incurred by using edge representations. Such approaches require both 1-edges and 0-edges (missing edges) to be provided as input, and as a consequence, approximate inference algorithms for these models usually require Ω(N2) time per iteration, precluding their application to larger real-world networks. In contrast, triangular modeling requires less computation, while providing equivalent or better inference quality. A triangular motif is a vertex triple containing 2 or 3 edges, and the number of such motifs is Θ(Σ i Di2) (where Di is the degree of vertex i), which is much smaller than N2 for low-maximum-degree networks. Using this representation, we develop a novel mixed-membership network model and approximate inference algorithm suitable for large networks with low max-degree. For networks with high maximum degree, the triangular motifs can be naturally subsampled in a node-centric fashion, allowing for much faster inference at a small cost in accuracy. Empirically, we demonstrate that our approach, when compared to that of an edge-based model, has faster runtime and improved accuracy for mixed-membership community detection. We conclude with a large-scale demonstration on an N ≈ 280, 000-node network, which is infeasible for network models with (N2) inference cost.",
author = "Qirong Ho and Junming Yin and Xing, {Eric P.}",
year = "2012",
language = "English (US)",
isbn = "9781627480031",
volume = "3",
pages = "2132--2140",
booktitle = "Advances in Neural Information Processing Systems",

}

TY - GEN

T1 - On triangular versus edge representations - Towards scalable modeling of networks

AU - Ho, Qirong

AU - Yin, Junming

AU - Xing, Eric P.

PY - 2012

Y1 - 2012

N2 - In this paper, we argue for representing networks as a bag of triangular motifs, particularly for important network problems that current model-based approaches handle poorly due to computational bottlenecks incurred by using edge representations. Such approaches require both 1-edges and 0-edges (missing edges) to be provided as input, and as a consequence, approximate inference algorithms for these models usually require Ω(N2) time per iteration, precluding their application to larger real-world networks. In contrast, triangular modeling requires less computation, while providing equivalent or better inference quality. A triangular motif is a vertex triple containing 2 or 3 edges, and the number of such motifs is Θ(Σ i Di2) (where Di is the degree of vertex i), which is much smaller than N2 for low-maximum-degree networks. Using this representation, we develop a novel mixed-membership network model and approximate inference algorithm suitable for large networks with low max-degree. For networks with high maximum degree, the triangular motifs can be naturally subsampled in a node-centric fashion, allowing for much faster inference at a small cost in accuracy. Empirically, we demonstrate that our approach, when compared to that of an edge-based model, has faster runtime and improved accuracy for mixed-membership community detection. We conclude with a large-scale demonstration on an N ≈ 280, 000-node network, which is infeasible for network models with (N2) inference cost.

AB - In this paper, we argue for representing networks as a bag of triangular motifs, particularly for important network problems that current model-based approaches handle poorly due to computational bottlenecks incurred by using edge representations. Such approaches require both 1-edges and 0-edges (missing edges) to be provided as input, and as a consequence, approximate inference algorithms for these models usually require Ω(N2) time per iteration, precluding their application to larger real-world networks. In contrast, triangular modeling requires less computation, while providing equivalent or better inference quality. A triangular motif is a vertex triple containing 2 or 3 edges, and the number of such motifs is Θ(Σ i Di2) (where Di is the degree of vertex i), which is much smaller than N2 for low-maximum-degree networks. Using this representation, we develop a novel mixed-membership network model and approximate inference algorithm suitable for large networks with low max-degree. For networks with high maximum degree, the triangular motifs can be naturally subsampled in a node-centric fashion, allowing for much faster inference at a small cost in accuracy. Empirically, we demonstrate that our approach, when compared to that of an edge-based model, has faster runtime and improved accuracy for mixed-membership community detection. We conclude with a large-scale demonstration on an N ≈ 280, 000-node network, which is infeasible for network models with (N2) inference cost.

UR - http://www.scopus.com/inward/record.url?scp=84877735494&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84877735494&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:84877735494

SN - 9781627480031

VL - 3

SP - 2132

EP - 2140

BT - Advances in Neural Information Processing Systems

ER -