Fig. 10

Computational times for sampling \(10\%\) of the Epinions (left), Slashdot (center) and Stanford (right) datasets. System sizes are increasing from left to right (Epinions and Slashdot have \(\sim 10^{4}\) nodes, but the latter has almost twice the number of edges, Stanford has \(\sim 10^{5}\) nodes and \(\sim 10^{6}\) edges). In the first two EXP sampling is not represented, as it took an average, respectively, of \(\approx 90s\) and \(\approx 400s\), and has not been performed for the last, due to excessive computational time. Boxplots represent 10 rounds of sampling