Table 1 Pseudocode of our algorithm for community detection via total correlation optimization

From: Leveraging multivariate information for community detection in functional brain networks

• compute TSE curve

• for it = 1 through it = ITER

     • plant a random seed partition C0

     • compute TCscore on C0 → score0

     • for h=0 through h=H

          • define temperature \(T={T}_{0}\times {(1-{h}_{frac}/H)}^{h}\)

          • switch module assignment for a random node i → Ci

          • compute new TCscore on Ci → scorei

          • if \(scor{e}_{i} > scor{e}_{0}\& \& \exp \left(\frac{scor{e}_{i}-scor{e}_{0}}{T}\right)\, > \, rand(0,1)\)

               • update partition  → C0 = C1

               • update TCscore → score0 = scorei

  1. We set the parameters as follows: H = 100,000 (number of iterations in the annealing process); ITER = 100 (number of times we perform the optimization through annealing); hfrac = 10; T0 = 1.