We propose a modification of MPGP algorithm for solving minimizing problem of strictly convex quadratic function subject to separable spherical constraints. This active set based algorithm explores the faces by the conjugate gradients and changes the active sets and active variables by the gradient projection with the Barzilai-Borwein steplength. We show how to use the algorithm for the solution of separable and equality constraints. The power of our modification is demonstrated on the solution of a contact problem with Tresca friction.
@article{702721,
title = {An optimal algorithm with Barzilai-Borwein steplength and superrelaxation for QPQC problem},
booktitle = {Programs and Algorithms of Numerical Mathematics},
series = {GDML\_Books},
publisher = {Institute of Mathematics AS CR},
address = {Prague},
year = {2013},
pages = {155-161},
url = {http://dml.mathdoc.fr/item/702721}
}
Pospíšil, Lukáš. An optimal algorithm with Barzilai-Borwein steplength and superrelaxation for QPQC problem, dans Programs and Algorithms of Numerical Mathematics, GDML_Books, (2013), pp. 155-161. http://gdmltest.u-ga.fr/item/702721/