Abstract
Graph embedding methods are becoming increasingly popular in the machinelearning community, where they are widely used for tasks such as nodeclassification and link prediction. Embedding graphs in geometric spaces shouldaid the identification of network communities as well, because nodes in thesame community should be projected close to each other in the geometric space,where they can be detected via standard data clustering algorithms. In thispaper, we test the ability of several graph embedding techniques to detectcommunities on benchmark graphs. We compare their performance against that oftraditional community detection algorithms. We find that the performance iscomparable, if the parameters of the embedding techniques are suitably chosen.However, the optimal parameter set varies with the specific features of thebenchmark graphs, like their size, whereas popular community detectionalgorithms do not require any parameter. So it is not possible to indicatebeforehand good parameter sets for the analysis of real networks. This finding,along with the high computational cost of embedding a network and grouping thepoints, suggests that, for community detection, current embedding techniques donot represent an improvement over network clustering algorithms.