Paper Info
Title
Ego-Splitting Framework: from Non-Overlapping to Overlapping Clusters.
Abstract
We propose ego-splitting, a new framework for detecting clusters in complex networks which leverage the local structures known as ego-nets (i.e. the subgraph induced by the neighborhood of each node) to de-couple overlapping clusters. Ego-splitting is a highly scalable and flexible framework, with provable theoretical guarantees, that reduces the complex overlapping clustering problem to a simpler and more amenable non-overlapping (partitioning) problem. We can scale community detection to graphs with tens of billions of edges and outperform previous solutions based on ego-nets analysis. More precisely, our framework works in two steps: a local ego-net analysis phase, and a global graph partitioning phase. In the local step, we first partition the nodes' ego-nets using a partitioning algorithm. We then use the computed clusters to split each node into its persona nodes that represent the instantiations of the node in its communities. Finally, in the global step, we partition the newly created graph to obtain an overlapping clustering of the original graph.
Year
DOI
Venue
2017
10.1145/3097983.3098054
KDD '17: The 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining Halifax NS Canada August, 2017
Keywords
Field
DocType
Overlapping clustering,ego-nets,large-scale graph algorithms
Data mining,Cluster (physics),Graph,Computer science,Artificial intelligence,Complex network,Partition (number theory),Graph partition,Cluster analysis,Machine learning,Scalability
Conference
ISBN
Citations 
PageRank 
978-1-4503-4887-4
6
0.46
References 
Authors
24
3
Name
Order
Citations
PageRank
Alessandro Epasto123617.08
Silvio Lattanzi272046.77
renato paes333136.45