HACIA UN NUEVO INDICADOR DE CALIDAD PARA SOLUCIONES A PROBLEMAS MULTI-OBJETIVO POR MEDIO DE PROGRAMACIÓN GENÉTICA

SANDOVAL REYES, CRISTIAN (2019) HACIA UN NUEVO INDICADOR DE CALIDAD PARA SOLUCIONES A PROBLEMAS MULTI-OBJETIVO POR MEDIO DE PROGRAMACIÓN GENÉTICA. Maestría thesis, Universidad Autónoma de Chihuahua.

[img]
Vista Previa
Text
Tesis (1).pdf

Download (1363Kb) | Vista Previa

Resumen

In the real world it is not uncommon to find an optimization problem with multiple objectives in conflict, and therefore there is not a single optimal solution but a set of optimal solutions. Multi-objective Genetic Algorithms have resulted in a promising tool for solving these types of problems. However, an important issue in multi-objective optimization is to evaluate and compare the quality of the solution sets generated by different multi-objective algorithms. A simple way to compare the quality of solution sets is through visualization, but visual comparison cannot quantify the difference between solution sets since this problem becomes harder as the number of objective functions increases. Quality indicators arise as a quantitative way of comparing solution sets. The most commonly used indicator, probably due to its mathematical properties, is the so-called hypervolume. Hypervolume is not only used to compare sets of solutions, but also as a selection criterion in the optimization process within these algorithms. It is important to note that coming with the exact of hypervolume is computationally expensive, meaning that the running time grows exponentially with respect to the number of objectives. The latter being one of its main shortcomings, especially when the number of objective functions is high. In this thesis work a new approach is proposed to approximate the hypervolume value using models trained with Genetic Programming. The results are very promising, showing that some of the models can approximate the hypervolume value with an accuracy of up to 99% with a speed 10 to 80 times faster than other competing algorithms used for the same task.

Tipo de Documento: Tesis (Maestría)
Palabras Clave: Multi-objective optimization, multi-objective genetic algorithms, multi-objective quality indicators, hypervolume indicator, Genetic Programming.
Clasificación temática: Q Science > QA Mathematics
Usuario Remitente: Admin Administrador del Respositorio
Depositado: 28 Jul 2020 17:54
Ultima Modificación: 28 Jul 2020 17:54
URI: http://repositorio.uach.mx/id/eprint/265

Actions (login required)

Ver Objeto Ver Objeto

Universidad Autonoma de Chihuahua