On the decrease of a quadratic function along the projected-gradient path
Electronic transactions on numerical analysis, Tome 31 (2008), pp. 25-29
The Euclidean gradient projection is an efficient tool for the expansion of an active set in the activeset-based algorithms for the solution of bound-constrained quadratic programming problems. In this paper we examine the decrease of the convex cost function along the projected-gradient path and extend the earlier estimate given by Joachim Sch$\ddot $oberl. The result is an important ingredient in the development of optimal algorithms for the solution of convex quadratic programming problems.
Classification :
65K05, 90C20, 49M10
Keywords: bound-constrained quadratic programming, Euclidean gradient projection, rate of convergence
Keywords: bound-constrained quadratic programming, Euclidean gradient projection, rate of convergence
@article{ETNA_2008__31__a21,
author = {Dost\'al, Zden\v{e}k},
title = {On the decrease of a quadratic function along the projected-gradient path},
journal = {Electronic transactions on numerical analysis},
pages = {25--29},
year = {2008},
volume = {31},
zbl = {1186.65074},
language = {en},
url = {http://geodesic.mathdoc.fr/item/ETNA_2008__31__a21/}
}
Dostál, Zdeněk. On the decrease of a quadratic function along the projected-gradient path. Electronic transactions on numerical analysis, Tome 31 (2008), pp. 25-29. http://geodesic.mathdoc.fr/item/ETNA_2008__31__a21/