This work presents a new new hybrid metaheuristic and bioinspired in nature, based on three main pillars, namely: Genetic Algorithms; Game Theory; and Fuzzy Logic. Thus, the Fuzzy Social Interaction Genetic Algorithm , or F-SIGA, is based and characterized by allowing individuals in the population the possibility to participate in the Social Interaction process. This step is prior to the selection process for the generation of offspring, and in it they can make gains through disputes with other individuals. For this, each individual is characterized by two chromosomes, one related to solving the problem in question; and, the other, with the gene encoding a behavioral strategy. As disputes environment is used the Prisoner's Dilemma game, in 2-person and N-person versions, including the fuzzy approach. In addition, individuals are evaluated by a fitness function that includes: a representation of the problem´ solution, the gains made in disputes and also the experience factor, using the experience acquired by the individual as well as a component to assist in the evolving process of the population. This characteristic gave rise to the ESIA algorithm - not originally planned - where only the information obtained from social interactions are considered in the selection of individuals for reproduction stage. Methodologically, the work evolved into the emergence of the ESIA algorithm, which is a new class of Evolutionary Algorithms based on social interaction. Thus, this work presents four algorithms: SIGA, NpSIGA, F-SIGA and ESIA, with its theoretical foundations and also practical results of applying them to global optimization problems with and without constraints; and the instances of the Travelling Salesman Problem.
|
