This node is located in the Systems > Algorithms > Optimization folder in the Modules pane.
The Unigene Evolutionary Algorithm (UEA) follows the general Evolutionary Algorithms principles described in Methods for Multi-Disciplinary Optimization and Robustness Analysis. UEA is a population-based optimization algorithm specialized for solving highly-constraint single-objective optimization applications. It comes with a tuned internal gene representation and combines multiple crossover and mutation operators for design variation. The objective and constraint definitions are formulated through a single scalar-valued objective (fitness) function to be minimized. This fitness function is a weighted sum of fitness terms, where each fitness term represents the rating for an objective or a constraint, respectively. The way objectives and constraints are mapped to fitness terms is unique to the UEA.
Node Settings
To access the options shown in the following tables, double-click the Unigene Evolutionary Algorithm (UEA) system on the Scenery pane and switch to the Settings tab.
The following General settings are available.
Initialization
| Option | Description |
|---|---|
| Population size | The number of individuals of the first population. |
| Type | The sampling method used to generate the first population, either uniform or Gaussian. |
Algorithm
| Option | Description |
|---|---|
| Number of generations | The maximum number of generations after which the optimization stops. |
| Crossover Probability | The probability of mating two individuals. |
| Mutation Probability | The probability of mutating an individual. |
| Tournament size | The number of individuals participating in each tournament. (Refers to the selection operator.) |
The objective and constraint definitions are formulated through a single scalar-valued objective (fitness) function to minimize. This fitness function is a weighted sum of fitness terms, where each fitness term represents the rating for an objective or a constraint.
For a given individual x, the fitness function is a weighted sum of fitness terms:
where each fitness term Di represents the rating for an objective or a constraint and ωi the corresponding weight. The following tables describe the additional parameters used to compute such ratings.
Objectives
| Option | Description |
|---|---|
| Initial Value | Initial value of the objective. Serves as a reference value. |
| Target Value | Estimated best-possible objective. |
| Weighting Factor | Relative weight of the corresponding fitness term. |
Constraints
| Option | Description |
|---|---|
| Feasible Tolerance | Tolerance for constraint values (beyond the limit value) still considered feasible for the problem at hand. |
| Weighting Factor | Relative weight of the corresponding fitness term |
The following Advanced options are available.
Crossover Operators
| Crossover Type | Option | Description |
|---|---|---|
| Uniform | Number of Genes | Number of genes to be swapped. |
| Uniform | Weight | Relative weight for the random selection of the crossover operator. |
| Intermediate | Swap probability | Independent probability for each gene to be swapped. |
| Intermediate | Weight | Relative weight for the random selection of crossover operator. |
| Segment | Optional parameter | Optional parameter to be added/subtracted from parent values before executing the operator. |
| Segment | Weight | Relative weight for the random selection of crossover operator. |
| Hypercube | Optional parameter | Independent probability for each gene to be swapped. |
| Hypercube | Weight | Relative weight for the random selection of crossover operator. |
Each time crossover between two individuals is applied, a crossover operator is randomly selected among the available ones according to the specified relative weights.
Mutation Operators
| Mutation Type | Option | Description |
|---|---|---|
| Uniform | Mutation probability | The probability that a gene mutates. |
| Uniform | Weight | Relative weight for the random selection of the mutation operator. |
| Deterministic uniform | Number of Genes | Number of genes to mutate in the individual. |
| Deterministic uniform | Weight | Relative weight for the random selection of mutation operator. |
| Gaussian | Mutation probability | The probability that a gene mutates. |
| Gaussian | Weight | Relative weight for the random selection of mutation operator. |
| Deterministic Gaussian | Number of Genes | Number of genes to mutate in the individual. |
| Deterministic Gaussian | Weight | Relative weight for the random selection of mutation operator. |
| Gaussian and Deterministic Gaussian | Sigma | Reference standard deviation for Gaussian-type mutation operators. For each gene, the actual standard deviation is obtained by multiplying the reference one to the corresponding variable’s range. |
Each time mutation is applied, a mutation operator is randomly selected among the available ones according to the specified relative weights.