This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
Benchmarking different clustering methods on simulated data.
JointCluster detected implanted clusters on instances of randomly generated graphs better than Coassociation and single tree methods, especially when the cluster structure was not strong, in two cases: (A) noise level in both were varied together, (B) noise level of was fixed at and of was varied from 0 to 16. The quality of the clustering detected by a method is measured as the standard Jaccard index measure between the detected and true clustering (y axis), averaged over all random instances for each setting of the noise level parameter (x axis). The average number of edges incident at each node is 16, so indicates a false positive rate of 50%.