Autors: POPOVICIU,N., BONCUT,M. Title: Algorithms of choosing the pre-conditioner matrix for quadratic programming with linear bound constraints Keywords: Quadratic programming, preconditioning algorithm, bound cons Abstract: The paper begins with several quadratic programming models (QP). For QP model with linear bound constraints a preconditioning technique [2] is necessary, in a neural network frame. This technique reduces the susceptibility of the system to errors. Two algorithms of preconditioning are presented: the first one is based on one matrix and the second is based on two matrixes. Both algorithms are used in three numerical applications. Each application ends by a test of correctitude of computations. References Issue
|
Вид: публикация в национален форум