A Modified Newton Method for Multilinear PageRank

作者:Guo, Pei-Chang*; Gao, Shi-Chen; Guo, Xiao-Xia
来源:Taiwanese Journal of Mathematics, 2018, 22(5): 1161-1171.
DOI:10.11650/tjm/180303

摘要

When studying the multilinear PageRank problem, a system of polynomial equations needs to be solved. In this paper, we propose a modified Newton method and develop a monotone convergence theory for a third-order tensor when alpha < 1/2. In this parameter regime, the sequence of vectors produced by the Newton-like method is monotonically increasing and converges to the solution. When alpha > 1/2 we present an always-stochastic modified Newton iteration. Numerical results illustrate the effectiveness of this method.