Publicação

A new algorithm to identify all global maximizers based on simulated annealing

Ver documento

Detalhes bibliográficos
Resumo:In this paper we consider the problem of finding all the global maximizers of a given nonlinear optimization problem. This type of problem appears, for example, in the phase-shift analysis of experimental data on scattering in nuclear and elementary particle physics and in a local reduction method for solving semi-infinite programming problems. The simulated annealing (SA) method is a stochastic method and it is well documented in the literature. Its most important property, as a global optimizer, is that asymptotic convergence to a global solution can be proven. However, in general, the SA algorithm finds just one global optimum. The function stretching technique carries out a two-step transformation of the objective function aiming to eliminate local optima while preserving the global ones. We propose a new algorithm which combines the SA algorithm with a function stretching, to generate a sequence of nonlinear maximization problems that are defined whenever a new maximizer is identified. To find all global maximizers, we apply the SA algorithm to this sequence of maximization problems. Results of numerical experiments with a set of well-known test problems in the global optimization literature show that the proposed method is effective. We also compare the performance of our algorithm with other multi-local solvers.
Autores principais:Pereira, Ana I.
Outros Autores:Fernandes, Edite M.G.P.
Assunto:Multi-global optimization Simulated annealing
Ano:2005
País:Portugal
Tipo de documento:documento de conferência
Tipo de acesso:acesso aberto
Instituição associada:Instituto Politécnico de Bragança
Idioma:inglês
Origem:Biblioteca Digital do IPB
Descrição
Resumo:In this paper we consider the problem of finding all the global maximizers of a given nonlinear optimization problem. This type of problem appears, for example, in the phase-shift analysis of experimental data on scattering in nuclear and elementary particle physics and in a local reduction method for solving semi-infinite programming problems. The simulated annealing (SA) method is a stochastic method and it is well documented in the literature. Its most important property, as a global optimizer, is that asymptotic convergence to a global solution can be proven. However, in general, the SA algorithm finds just one global optimum. The function stretching technique carries out a two-step transformation of the objective function aiming to eliminate local optima while preserving the global ones. We propose a new algorithm which combines the SA algorithm with a function stretching, to generate a sequence of nonlinear maximization problems that are defined whenever a new maximizer is identified. To find all global maximizers, we apply the SA algorithm to this sequence of maximization problems. Results of numerical experiments with a set of well-known test problems in the global optimization literature show that the proposed method is effective. We also compare the performance of our algorithm with other multi-local solvers.