Abstract:
Some features of the numerical implementation of solving tree-dimensional
Fredholm integral equations of the first kind for vector functions with
application of multiprocessor systems are considered. The Tikhonov regularization
is applied to solve this ill-posed problem. The conjugate gradient method is used
as a minimization procedure. The choice of the regularization parameter is
performed according to the generalized discrepancy principle. A parallelization
scheme for this problem is proposed; the efficiency of the approach under
consideration is shown by the example of restoring magnetization parameters.
This work was supported by the Russian Foundation for Basic Research
(projects 08-01-00160-a and 10-01-91150-NFSC). The numerical results were
obtained using the Computing Cluster of Moscow State University.
Keywords:three-dimensional Fredholm integral equations of the first kind; conjugate gradient method; Tikhonov regularization; parallel algorithms.