High-performance implementations of a clustering algorithm for finding network communities

The size and interconnectedness of social networks continues to increase. As a result, finding communities or subsets of like nodes within these large networks has become a resource-intensive endeavor. In this paper, we characterize community-finding organized on the basis of network/set properties, and describe an agglomerative algorithm called egocentric community finding. The primary contribution of this paper is a performance evaluation in which the egocentric data-mining algorithm is optimized for execution on various computing platforms, including GPU's, multicore systems, and large-scale distributed systems.