Robustness & reliability

Real network data are noisy and incomplete, and Infomap always returns some partition. The techniques here keep sparse or under-sampled data from fragmenting into small, spurious modules.

• Incomplete data and regularization: when links are missing, standard Infomap overfits into many small spurious modules. The Bayesian regularized map equation adds a prior network so that only well-supported communities survive.