Network Sampling Based on NN Representatives

Abstract: The amount of large-scale real data around us increase in size very quickly and so does the necessity to reduce its size by obtaining a representative sample. Such sample allows us to use a great variety of analytical methods, whose direct application on original data would be infeasible. There are many methods used for different purposes and with different results. In this paper we outline a simple and straightforward approach based on analyzing the nearest neighbors (NN) that is generally applicable. This feature is illustrated on experiments with weighted networks and vector data. The properties of the representative sample show that the presented approach maintains very well internal data structures (e.g. clusters and density). Key technical parameters of the approach is low complexity and high scalability. This allows the application of this approach to the area of big data.