Editorial: Scalable Network Generation & Analysis

Philippe J. Giabbanelli, Samarth Swarup, Renaud Lambiotte, Giuseppe Mangioni

Abstract: Innovative practices for large-scale network generations are proposed by Alam and Perumalla. The authors produced a generator that can realize any desired degree distribution, thus making it a particularly flexible tool for common tasks such as generating equivalent random networks (e.g., to test the presence of a property in an empirical network) or creating a synthetic population. Most importantly, the generator can achieve rates of over 50 billion edges per second through its high utilization of a single modern GPU. By creating the first GPU-based algorithm to generate networks with a given degree distribution, this work is also an invitation for the network science community to explore numerous potential extensions, such as using multiple GPUs (on the same machine and/or via an interconnection network) to realize even larger networks in a timely manner.The creation of a network generator is often accompanied by a characterization of its instances based on the generator's parameters. This task can be arduous as the apparent simplicity of a generator's rules can lead to highly intricate instances once these rules are applied repeatedly, particularly in the presence of stochasticity or bifurcations. The cornerstone of the novel approach offered by Murase et al. consists of creating a massive number of instances from the generators and to use them as training data for a machine learning model that is then able to predict an instance's characteristic from the parameter valu...