Ouvrir le document
|
Informations générales |
Auteur |
Färber, Markus; Brüderlin, Beat |
Publié |
2010
|
Edition |
|
Extension |
|
ISBN |
|
Abstract |
This work presents a novel approach for multivariate root finding combining
search space decomposition by generalised quad-trees with an improved
variant of the secant method. The search space is decomposed
adaptively to find promising start points for the local search.
To use all the available information generated by the decomposition
into rectangular cells, the Newton-Raphson iteration featuring linearisation
of the function has been improved by non-linear Lagrange interpolating
polynomials.
The method has been developed in the context of geometric constraint
solving, but is applicable to a wider range of problems. |
|