A parallel genetic algorithm in multi-objective optimization

Parallel Multi-Objective Evolutionary Algorithms: A Comprehensive Survey

“…In 2009, Wang and Ju [53] proposed an island model version of NSGA-II denoted as PNSGA. PNSGA employs only two self-evolving islands: an elite population island (EP) that stores the global non-dominated individuals and the search population island (SP) whose purpose is to explore the search space.…”

An Improved Nonlinear Multi-Objective Optimization Problem Based on Genetic Algorithm

“…non-Pareto techniques including : polymerization method , VEGA algorithm lexicographic method , G-constraint method , the target vector method ; 363 In the algorithm design, Yu-Ping Wang [8], respectively, the orthogonal design, uniform design combined with genetic algorithm for solving multi -objective optimization gives the new method. [9] And so introduces a ε -constraint method Augmented Lagrange algorithm for multi-target collaboration.…”

“…(b) The initial population: For the generation of the initial population, using a random generator of. (c) Fitness function: fitness function using equation (8) in the design of the fitness function. (d) Selection operator: Use roulette wheel selection and elitist strategy of combining strategy with a parent in the fitness value of the top surface of M individuals through crossover and mutation to replace the population after the fitness value of row M individuals in the final, so that offspring and parent to participate in competition, in order to prevent destruction of crossover and mutation operations populations fine mode.…”

Parallel Single and Multiple Objectives Genetic Algorithms

“…Coarse grained parallel GAs are possibly the most popular model. Many papers have been published on course-grained parallel GAs (Arryo & Conejo, 2002;Zhi-xin & Gang, 2009;Xu, Ge, & Ming, 2009;Yan, Jumin, & Zhuoshang, 2000). Sometimes these models are known as distributed GAs as they are usually implemented on distributed memory MIMD computers.…”

“…MOGA are highly suitable for parallelization as crossover and mutation and in particular the time consuming fitness evaluation can be performed independently on different processors (Zhi-xin & Gang, 2009;Hiroyasu, Miki, & Watanabe, 2000). The main problem is the parallelization of the selection operator, where global information is required to determine the relative performance of an individual with respect to all others in the current population.…”