Evolutionary spectral clustering with adaptive forgetting factor

Xu, Kevin S.; Kliger, Mark; Hero, Alfred O.

doi:10.1109/icassp.2010.5495655

Cited by 17 publications

(11 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Another example of snapshot-based approach is based on state-of-the-art spectral clustering extended by consistency constraints [17] or adaptive forgetting factor [62]. A comprehensive study of evolutionary clustering can be found in a survey by M. Spiliopoulou [54].…”

Section: Snapshot-based Community Detectionmentioning

confidence: 99%

Discovering Dynamic Communities in Interaction Networks

Rozenshtein

Tatti

Gionis

2014

Machine Learning and Knowledge Discovery in Databases

View full text Add to dashboard Cite

Very often online social networks are defined by aggregating information regarding the interaction between the nodes of the network. For example, a call graph is defined by considering an edge for each pair of individuals who have called each other at least once -or at least k times. Similarly, an implicit social network in a social-media site is defined by considering an edge for each pair of users who have interacted in some way, e.g., have made a conversation, commented to each other's content, etc. Despite the fact that this type of definitions have been used to obtain a lot of insights regarding the structure of social networks, it is obvious that they suffer from a severe limitation: they neglect the precise time that the interaction between network nodes occurs.In this thesis we propose to study interaction networks, where one considers not only the underlying topology of the social network, but also the exact time instances that nodes interact. In an interaction network an edge is associated with a time stamp, and multiple edges may occur for the same pair of nodes. Consequently, interaction networks offer a more fine-grained representation that can be used to reveal otherwise hidden dynamic phenomena in the network.In the context of interaction networks, we study the problem of discovering communities, which are dense in terms of the underlying network structure, and whose edges occur in short time intervals. Such communities represent groups of individuals who interact with each other in some specific time instances, for example, a group of employees who work on a project and whose interaction intensifies before certain project milestones. We prove that the problem we define is NP-hard, and we provide effective algorithms by adapting techniques used to find dense subgraphs. We perform extensive evaluation of the proposed methods on synthetic and real datasets, which demonstrates the validity of our concepts and the good performance of our algorithms.

show abstract

Section: Snapshot-based Community Detectionmentioning

confidence: 99%

Discovering Dynamic Communities in Interaction Networks

Rozenshtein

Tatti

Gionis

2014

Machine Learning and Knowledge Discovery in Databases

View full text Add to dashboard Cite

show abstract

“…If ε is set too large, for example 0.50, the model might incur "over-smoothing", i.e. being unable to adapt to changes in the community structure on this period [42].…”

Section: Mit Social Evolution Datasetmentioning

confidence: 99%

A mathematical programming approach for sequential clustering of dynamic networks

Silva¹,

Bennett

Papageorgiou

et al. 2016

Eur. Phys. J. B

View full text Add to dashboard Cite

Abstract.A common analysis performed on dynamic networks is community structure detection, a challenging problem that aims to track the temporal evolution of network modules. An emerging area in this field is evolutionary clustering, where the community structure of a network snapshot is identified by taking into account both its current state as well as previous time points. Based on this concept, we have developed a mixed integer non-linear programming (MINLP) model, SeqMod, that sequentially clusters each snapshot of a dynamic network. The modularity metric is used to determine the quality of community structure of the current snapshot and the historical cost is accounted for by optimising the number of node pairs co-clustered at the previous time point that remain so in the current snapshot partition. Our method is tested on social networks of interactions among high school students, college students and members of the Brazilian Congress. We show that, for an adequate parameter setting, our algorithm detects the classes that these students belong more accurately than partitioning each time step individually or by partitioning the aggregated snapshots. Our method also detects drastic discontinuities in interaction patterns across network snapshots. Finally, we present comparative results with similar community detection methods for time-dependent networks from the literature. Overall, we illustrate the applicability of mathematical programming as a flexible, adaptable and systematic approach for these community detection problems.

show abstract

“…However, Ψ t is unknown in real applications so the goal is to estimate it as accurately as possible. If we take the estimate to be the convex combination defined in (1), it was shown in [13] that the optimal choice of α t that minimizes the MSE in terms of the Frobenius norm E…”

Section: Adaptive Evolutionary Clusteringmentioning

confidence: 99%

“…The algorithm involves computing the eigenvectors corresponding to the k largest eigenvalues of a normalized version ofW t , then discretizing the eigenvectors so that each node is assigned to a single community. We refer readers to [13] for additional details on the adaptive evolutionary spectral clustering algorithm.…”

Section: Adaptive Evolutionary Clusteringmentioning

confidence: 99%

“…Evolutionary clustering combines data from multiple time steps to compute the clustering result at a single time step, which allows clustering results to vary smoothly over time. Xu et al [13] recently proposed an evolutionary clustering framework that adaptively estimates the optimal weighted combination of current and past data to minimize a mean-squared error (MSE) criterion. We describe the framework in the following.…”

Section: Adaptive Evolutionary Clusteringmentioning

confidence: 99%

See 1 more Smart Citation

Tracking Communities in Dynamic Social Networks

Kliger

Hero

2011

Social Computing, Behavioral-Cultural Modeling and Prediction

Self Cite

View full text Add to dashboard Cite

Abstract. The study of communities in social networks has attracted considerable interest from many disciplines. Most studies have focused on static networks, and in doing so, have neglected the temporal dynamics of the networks and communities. This paper considers the problem of tracking communities over time in dynamic social networks. We propose a method for community tracking using an adaptive evolutionary clustering framework. We apply the method to reveal the temporal evolution of communities in two real data sets. In addition, we obtain a statistic that can be used for identifying change points in the network.

show abstract

Evolutionary spectral clustering with adaptive forgetting factor

Cited by 17 publications

References 10 publications

Discovering Dynamic Communities in Interaction Networks

Discovering Dynamic Communities in Interaction Networks

A mathematical programming approach for sequential clustering of dynamic networks

Tracking Communities in Dynamic Social Networks

Contact Info

Product

Resources

About