Ponente: Lavinia Ferariu, Associate Professor at the “Gheorghe Asachi”
Technical University of Iasi, Rumanía.
Fecha/hora: Miércoles 26 de septiembre, 11:30 horas.
Lugar: Seminario del Departamento de Informática e Ingeniería de Sistemas, edificio Ada Byron (Campus Río Ebro).
Resumen:
Genetic algorithms with Pareto ranking could be successfully applied to multiobjective optimisations. One of the main challenge stays in providing an unbiased rank assignment, which allows a progressive combination between decision and search, while keeping a good diversity of the best (nondominated) solutions. In this regard, the talk presents how dominance analysis can be used for partially sorting the solutions in multiobjective genetic algorithms and gives a short overview of the most popular algorithms in the field. Then, the presentation continues with some advanced techniques proposed for integrating the genetic search with the preference for the middle of the best Pareto fronts. These techniques consider an adaptive labelling of the preferred solutions, according to the layout of fronts. The main targeted outcome is obtaining better nondominated final solutions, well distributed in the areas which are usually preferred by the practitioners. The exemplifications refer to applications like robot path planning and system identification.
