A variant of the Power–Arnoldi algorithm for computing PageRank

Abstract For computing PageRank problems, a Power–Arnoldi algorithm is presented by periodically knitting the power method together with the thick restarted Arnoldi algorithm. In this paper, by using the power method with the extrapolation process based on trace (PET), a variant of the Power–Arnoldi algorithm is developed for accelerating PageRank computations. The new method is called Arnoldi-PET algorithm, whose implementation and convergence are analyzed. Numerical experiments on several examples are used to illustrate the effectiveness of our proposed algorithm.