Big Data Analytics 2013

Invited talk

Neighborhood Sampling for Estimating Local Properties on a Graph Stream

Srikanta Tirthapura

We consider the estimation of local graph properties, which concern subgraphs that lie within the neighborhood of a vertex, such as counting the number of cliques with a certain number of vertices. We present a new algorithm for sampling the edges of the graph, called “neighborhood sampling”, which works in a single pass through the edges of the graph, presented in an arbitrary order. The algorithm is practical and easy to implement.