The authors explore pair-matching algorithms in stochastic block models to optimize edge discovery within communities.
Sublinear regret achievable in pair-matching with underlying structure.