Numerical procedure for solving a minimization eigenvalue problem (Q1115097): Difference between revisions

A numerical procedure for solving the left eigenvalue problem \(\lambda +L(k)=\min_ j[L(j)+K(j,k)],\) and its right counterpart, is described; K is a (known) real \(n\times n\) matrix. The first stage uses a modification of the procedure studied by \textit{R. M. Karp} [Discr. Math. 23, 309-311 (1978; Zbl 0386.05032)] and \textit{M. v. Golitschek} [Numer. Math. 39, 65- 84 (1982; Zbl 0541.65009)] for finding the minimum cyclic average \(\lambda\) of K along with a minimizing cycle. The second stage employs this information to generate L, by a very similar algorithm.