Scaling algorithms for matrices
Abstract
We present an iterative algorithm, called SCALGM, which asymptotically scales both rows and columns of any given matrix such that each element of the scaled matrix is in the interval [-1, 1] and the elements of minimum magnitude are maximized. The object is to make the condition number reasonably small, thus causing the pivoting process in Gaussian elimination to work well, and to diagnose any instability in the elimination process. Numerical evidence is presented showing the effectiveness of the algorithm.
Collections
- OSU Dissertations [11222]