Configuring the Multi-Island Genetic Algorithm Technique

From the Optimize view in the Predict page, select Optimization Settings .
From the Optimization Technique options, select Multi-Island Genetic Algorithm Technique.

Enter or select the following:

Option	Description
Sub-Population Size	The size of each island population. The total population is spread out equally between the islands. The size of the total population depends on the number of islands and the size of the sub-population. The value type is integer. The default value is 10. Other possible values are $\geq 1$ .
Number of Islands	The number of islands. The value type is integer. The default value is 10. Other possible values are $\geq 1$ .
Number of Generations	The number of generations that will be evaluated by the algorithm. Each generation includes sub-populations on all islands. The value type is integer. The default value is 10. Other possible values are $\geq 1$ .
Rate of Crossover	The percentage of designs (expressed as a fraction) in each generation that will be subjected to the “crossover” operation. The value type is real. The default value is 1.0. Other possible values are $\geq 0$ and $\leq 1.0$ .
Rate of Mutation	The probability of mutation of every gene (0 or 1 character within a chromosome) of every individual. Mutation changes the value of the gene to the opposite (0 becomes 1, and vice versa). The value type is real. The default value is 0.01. Other possible values are $\geq 0$ and $\leq 1.0$ .
Rate of Migration	The percentage of designs (expressed as a fraction) of each island’s population that will be migrated to another island when migration occurs. The value type is real. The default value is 0.5. Other possible values are $\geq 0$ and $\leq 0.5$ .
Interval of Migration	The number of generations between each migration. The default value is 5.
Elite Size	The number of best individuals carried over from the parent generation to the child generation in each sub-population. The value type is integer. The default value is 1. Other possible values are $\geq 1$ .
Relative Tournament Size	Percentage of designs (expressed as a fraction of the sub-population) randomly selected from each sub-population, from which the best individual is then selected for the child generation. Tournament selection allows for duplicate copies of individuals to be selected for the child generation. The tournament size cannot be below 1; therefore, if the relative tournament size is too small, the resulting size of the design subset (tournament) adjusts to at least 1. A tournament size of 1 means that the child generation is selected randomly: first one design is selected randomly to create the tournament, then the best (the only one) individual is selected from the tournament. The maximum tournament size is equal to the size of the sub-population. In this case the child generation will consist entirely of duplicates of the best individual from the parent generation. The value type is real. The default value is 0.5. Other possible values are $\geq 0$ and $\leq 1.0$ .
Penalty Base	Multi-Island Genetic Algorithm evaluates the quality of a design point using the combined value of the objective function and penalty function. When calculating the penalty function of the design, the Penalty Base option can be used for all designs that violate at least one constraint. This option allows you to better differentiate feasible designs with a slightly higher objective function from infeasible designs with a slightly lower objective function. The total penalty function is calculated as follows: Penalty $=$ Penalty Base $+$ Penalty Multiplier $\times$ Sum (VIOLATION_i $\times$ W_i / S_i) ^ Penalty Exponent where `Violation`_i is the `i – th` constraint violation value, `W`_i is the corresponding weight factor, and `S`_i is the corresponding scale factor. The PenaltyBase option is set to zero if no constraints are violated. The default value is 0.0.
Penalty Multiplier	The Penalty Multiplier is used to increase or decrease the effect of the total constraint violations on the measure of the design quality. The default value is 1000.0.
Penalty Exponent	The Penalty Exponent can be used to increase or decrease the nonlinearity of the effect of the total constraint violations on the penalty function value. The type of value is integer. The default value is 2.
Maximum Failed Runs	The maximum number of failed subflow evaluations that can be tolerated by the optimization technique. If the number of failed runs exceeds this value, the optimization adapter will terminate execution. To disable this feature, set this option to any negative value (for example, –1). When this option is set to a negative value, the optimization will continue execution despite any number of failed subflow runs.
Failed Run Penalty Value	The value of the Penalty parameter that is used for all failed subflow runs. The default value is 1.0E30.
Failed Run Objective Value	The value of the Objective parameter that is used for all failed subflow runs. The default value is 1.0E30.
Default Variable Bound (Absolute Value)	The upper and lower bound for all variables without bounds. For the lower bound, this value will be multiplied by –1 (negative value). It is important for the Multi-Island Genetic Algorithm to have bounds for all variables because the algorithm works by dividing the range of each variable into a large number of steps.
Use fixed random seed	If this option is selected, the random number generator used by the optimization algorithm is seeded using the value specified in the corresponding text box. All executions of the Optimization adapter will use the same sequence of random numbers and, therefore, will produce the same design points. This arrangement is useful for debugging the optimization process when it is necessary to reproduce the same sequence of design points. If this option is not selected, the random number generator is seeded by using the clock time at the moment of execution.

Click OK to save your changes and to close the Optimization Editor.