Divides the population into several islands. Performs traditional genetic operations on each island separately, then migrates individuals between the islands. Searches many designs and multiple locations of the design space.
In the Multi-Island Genetic Algorithm (MIGA), like other genetic algorithms, each design point is perceived as an individual with a certain value of fitness based on the value of the objective function and constraint penalty. An individual with a better value of the objective function and penalty has a higher fitness value. Each individual is represented by a chromosome in which the values of design variables are converted into a binary string of 0 and 1 characters. This conversion is called "encoding" of the individual. Each population of individuals (a set of design points) is altered via the genetic operations of "selection", "crossover", and "mutation". Each design of a population is then evaluated and its fitness value is determined. A new population of designs is selected from the original set of designs: a process based on a survival of the fittest scheme. New designs are created by the genetic crossover operation: chromosomes of two individuals are crossed at 2 points and the genes between those points are swapped in the two chromosomes resulting in two new individuals. Genetic operation of mutation changes a value of a randomly selected gene in a chromosome to further increase the variability of the population and avoid stagnation in the evolution process. Multi-Island Genetic Algorithm preserves the best individuals from the previous generation without alteration. This operation is called "elitism". Elitism guarantees that the best genetic material is carried over to the child generation.
The selection operation in Multi-Island Genetic Algorithm employs the so-called "tournament selection" scheme. In the tournament selection, the best individuals are selected not from the whole population, but rather from a smaller subset of randomly selected individuals. This scheme allows for duplicate individuals in the child population. The size of the subset from which each best individual is selected is calculated using the value of the relative tournament size. Reducing the relative tournament size will increase the randomness in the selection process. Increasing the tournament size will result in more duplicates of the best individuals in the child population.
The main feature of Multi-Island Genetic Algorithm that distinguishes it from traditional genetic algorithms is the fact that each population of individuals is divided into several sub-populations called "islands". All traditional genetic operations are performed separately on each sub-population. Some individuals are then selected from each island and migrated to different islands periodically. This operation is called "migration". Two parameters control the migration process: migration interval which is the number of generations between each migration, and migration rate which is the percentage of individuals migrated from each island at the time of migration.