Perron vector optimization applied to search engines (Q2448379)

This paper is concerned with the study of optimizing the Perron eigenvector of a controlled matrix. In the first part of the paper, a scalable algorithm for the local optimization of a scalar function of the Perron eigenvector over a set of nonnegative irreducible matrices is introduced. Next, the author proposes an optimization algorithm which couples gradient and power iterations. Finally, he presents some applications of Perron vector optimization to Kleinberg's HITS algorithm and to Tomlin's HOTS algorithm (e.g. optimization of scalar functions regarding authority from the first and HOTS scores from the latter).