A Useful Method to Improve Convergence in Least-Squares Fitting Procedures
J. Oswald, Helmut Feichtinger, R. Czaputa · physica status solidi (a) · 1984
An easy-to-handle method is presented avoiding problems as to convergence and/or processing time in cases where parameters of strongly differing weight enter the function by which an experimental data series is to be reproduced. The method consists basically of a flattening of strong relative minima in the mean square deviation and may be easily inserted as a subroutine within a fitting algorithm.