A Sensitivity-Analysis Approach for Domain Partitioning in Multisplitting Algorithms

VAZQUEZ, G.E.; BRIGNOLE, N.B.

Proceedings WCCM

Springer Verlag

Año: 2002; p. 1 - 1

A partitioning strategy that accelerates the convergence of parallel-distributed multisplittingalgorithms for nonlinear constrained optimization is proposed in this work. The technique is based on a sensitivity analysis of the optimization variables. It is shown that the way in which those variables are grouped affects algorithmic performance. Several industrial test problems were employed to assessthe scope and capabilities of the technique. It was determined that the method succeeds in reducing the amount of iterations as well as the number of function evaluations, which implies improvements in run times that become more significant as problems grow in size.