Institute of Theoretical Informatics, Algorithmics

Static Graph Generators

Among other tasks, we use systematic batches of random generated instances for the evaluation of clustering algorithms. Arguably, the relevance or realism of many random models such as Erdos-Renyi-type graphs can be questioned, however, the possibility to very clearly scale, parameterize and predict properties of such graphs render them valuable to understanding the behavior of an algorithm.

In short, our preclustered random graph generator is a highly efficient static version of our dynamic generator. Its parameters resemble those of the dynamic case. Its written in C++ and can be downloaded here:

static G(n,pin,pout) generator (alpha version)

This generator comes with no manual or description for now, as it is largely self-explanatory. Calling Gnpinpout with no parameters lists possible parameters. The generator itself is a class that can be used e.g., within a class running an evaluation, iterates over parameters and tests for statistical significance. The class Gnpinpout.cpp is a wrapper which exemplifies how command line calls can be passed on to the generator. In fact Gnpinpout is a ready-to-use wrapper for manual or script usage - however, if you do use scripts, take good care to pass on a random value to it, otherwise you might end up with the same network a hundred times (if its small). Other generators will follow.