Multistart coupled with a derivative-free filter local search for locating mult...

• Author(s): Fernandes, Florbela P. ; Pereira, Ana I. ; Costa, Maria F.P. ; Fernandes, Edite M.G.P.
• Date: 2012
• Persistent ID: http://hdl.handle.net/10198/7483
• Origin: Biblioteca Digital do IPB
• Subject(s): Multistart; Derivatives-free
• Document Type: Conference Object
• Language: English

Description

A multistart technique coupled with a derivative-free lter local search algorithm for locating all the optimal solutions of a nonconvex constrained optimization is presented. To reach a fast convergence to the optimal solutions, the local search procedure is based on descent directions. The lter-set concept is introduced to handle the constraints of the problem. The generated direction vector is descent for the objective function if the sample point is feasible; otherwise, it is descent for the constraint violation. Numerical experiments with benchmark problems are reported and a comparison with other stochastic methods is included.