1: \begin{abstract}

2: We study a classical iterative algorithm for balancing matrices in the $L_\infty$ norm via a scaling transformation. This algorithm, which goes back to Osborne and Parlett \& Reinsch in the 1960s,

3: is implemented as a standard preconditioner in many numerical linear algebra packages.

4: Surprisingly, despite its widespread use over several decades, no bounds were known on

5: its rate of convergence.  In this paper we prove that, for any irreducible $n\times n$ 

6: (real or complex) input matrix~$A$, a natural variant of the algorithm converges in 

7: $O(n^3\log(n\rho/\varepsilon))$ elementary balancing operations, where $\rho$ measures

8: the initial imbalance of~$A$ and $\varepsilon$ is the target imbalance of the output matrix.  

9: (The imbalance of~$A$ is $\max_i |\log(\aiout/\aiin)|$, where $\aiout,\aiin$ are the maximum

10: entries in magnitude in the $i$th row and column respectively.)  This bound is tight

11: up to the $\log n$ factor.  A balancing operation

12: scales the $i$th row and column so that their maximum entries are equal, and requires

13: $O(m/n)$ arithmetic operations on average, where $m$ is the number of non-zero elements in~$A$.

14: Thus the running time of the iterative algorithm is $\Otilde(n^2m)$. This is the first time bound

15: of any kind on any variant of the Osborne-Parlett-Reinsch algorithm.

16: We also prove a conjecture of Chen that characterizes those matrices for which

17: the limit of the balancing process is independent of the order in which 

18: balancing operations are performed.

19: \end{abstract}

20: