Fig. 2
From: Markov chain Monte Carlo for active module identification problem
Ranking AUC values for simulated instances for graphs on a 100, 500 and 2,034 vertices with different parameters of beta-uniform mixture. The proposed MCMC method is compared with the following methods: ranking by input p-values (left), semi-heuristic ranking from [12] (middle) and BioNet-based ranking (right). One arrow corresponds to one experiment. The head of each arrow points to the AUC score of the MCMC method, while the tail indicates the AUC score of the corresponding alternative method. Thus, upward arrows indicate instances in which the MCMC method has a higher AUC score. Dots indicate the instances for which the results are equal. Color depends on which method has better AUC