Abstract
A new method for minimising a sum of squares of non-linear functions is described and is shown to be more efficient than other methods in that fewer function values are required.