Multilinear PageRank: Uniqueness, error bound and perturbation analysis

Abstract In this paper, we revisit the multilinear PageRank problem. Under the framework of tensor, we establish several new and tighter uniqueness conditions for the multilinear PageRank vector. Meanwhile, a refined error bound for the inverse iteration as well as the new perturbation bounds under different norms, which improve the existing ones in the current literature, are developed with feasible computations. Several numerical examples are given to validate the significant effectiveness of the proposed bounds.