Two-phase algorithm for global optimization

Author(s)

Gabriel Haeser

• Document type: Master's Dissertation
• Press: Campinas, SP.
• Institution: Universidade Estadual de Campinas (UNICAMP). Instituto de Matemática, Estatística e Computação Científica
• Defense date: 2006-09-03

Examining board members

Márcia Aparecida Gomes Ruggiero; José Mario Martínez Pérez; Ernesto Julián Goldberg Birgin

• Advisor: Márcia Aparecida Gomes Ruggiero

Abstract

In this work we study the theory behind some classical heuristics for global optimization, and a generalization of genetic algorithms from Aarts, Eiben and van Hee. We propose an algorithm for global optimization of box-constrained differentiable problems, using simulated annealing and the local solver GENCAN. Numerical experiments are presented for the OVO problem (Order-Value Optimization) and 28 classical problems. For general nonlinear programming problems, we mention some ideas of how to use local solvers and global heuristics towards good algorithms for global optimization, we also propose an algorithm based on simulated annealing with local solver ALGENCAN (AU)