1: \begin{abstract}

2:   It is shown that the problem of balancing  a nonnegative matrix by positive diagonal matrices can be

3:   recast as a constrained nonlinear  multiparameter eigenvalue problem.  Based on this equivalent formulation

4:   some adaptations of the  power method and  Arnoldi process are proposed for  computing the  dominant eigenvector which

5:   defines the structure of the diagonal transformations. Numerical results illustrate

6:   that our novel  methods  accelerate significantly  the convergence of the  customary Sinkhorn-Knopp iteration

7:  for matrix balancing  in the case  of clustered dominant eigenvalues. 

8: \end{abstract}

9: