Two-Phase Methods for Global Optimization

Abstract

Virtually all methods for global optimization consist of two phases: a global phase, aimed at thorough exploration of the feasible region or subsets of the feasible region where it is known the global optimum will be found, and a local phase aimed at locally improving the approximation to some local optima. Often these two phases are blended into the same algorithm, which automatically switches between exploration and refinement. In this paper, some methods will be reviewed in which the two phases are well separated. A formal definition of two-phase methods will be given and some of the characteristics of these methods will be outlined.