zbMATH — the first resource for mathematics

The use of the $$L$$-curve in the regularization of discrete ill-posed problems. (English) Zbl 0789.65030
The authors elaborate a new strategy for the choice of the regularization parameter in regularization methods for discrete linear models $$Kx = y$$ of ill-posed problems where $$K$$ is a matrix of dimension $$m\times n$$, $$m\geq n$$. The choice is based on the behaviour of the so called $$L$$- curve, i.e. the plot of the size of the regularized solution versus the size of the corresponding residual. This choice is related to the generalized cross validation but it can be implemented more efficiently.
Reviewer: G.Vainikko (Tartu)

MSC:
 65F20 Numerical solutions to overdetermined systems, pseudoinverses
Full Text: