ε Suboptimality Based Accelerated Termination for Quadratic Programming Using Dual Decomposition
Résumé
In solving quadratic programming, analytical and dual decomposition based iterative method are 2 main approaches. However, the defects are evident: the former only works with small size problems and the latter only guarantees feasibility in the limit of iterations. In this paper, we propose a proactive method by combining these 2 methods to solve the optimal solution through dynamically identifying the active inequalities constraints. Further, to faster terminate the iterative process, we propose a suboptimal method based on cone programming to deliver feasible solutions with suboptimality guarantee. In addition to the mathematical proofs provided, various random simulations illustrate the effectiveness of the suboptimal method.
Origine : Fichiers produits par l'(les) auteur(s)