BDBComp
Parceria:
SBC
Automatic Design of Algorithms Applied to the Multi-Objective TSP Problem

Thainá MarianiGiovani GuizzoSilvia R. VergilioAurora T. R. Pozo

Research on Multi-Objective Evolutionary Algorithms (MOEAs) has grown in the last years. As a consequence, a great number of algorithms has been proposed. Therefore, for solving a problem, it is necessary to choose an algorithm and select its parameters and components. This can be a difficult task for non-expert practitioners. The use of Grammatical Evolution (GE) can help in this task allowing the automatic generation of custom MOEAs. GE has been used for the automatic generation of mono-objective algorithms, however, for MOEAs this is an underexplored subject. Considering this fact, this work proposes an approach based on GE for the automatic design of MOEAs. Using six instances of the Multi-Objective Traveling Salesman problem the approach is empirically evaluated, and the generated algorithms are compared to three well- known MOEAs. The results show that the proposed approach is able to generate novel algorithms that outperform the compared algorithms in all instances of the problem being solved.

http://www.lbd.dcc.ufmg.br/colecoes/eniac/2016/037.pdf

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato: bdbcomp@lbd.dcc.ufmg.br
     Mantida por:
LBD