Termination Methods

Termination is the criterion by which the genetic algorithm decides whether to continue searching or stop the search. Each of the enabled termination criterion is checked after each generation to see if it is time to stop.

OptiGen Library includes the following types of termination:

Generation Number - A termination method that stops the evolution when the user-specified max number of evolutions have been run. This termination method is always active.

Evolution Time - A termination method that stops the evolution when the elapsed evolution time exceeds the user-specified max evolution time. By default, the evolution is not stopped until the evolution of the current generation has completed, but this behavior can be changed so that the evolution can be stopped within a generation.

Fitness Threshold - A termination method that stops the evolution when the best fitness in the current population becomes less than the user-specified fitness threshold and the objective is set to minimize the fitness. This termination method also stops the evolution when the best fitness in the current population becomes greater than the user-specified fitness threshold when the objective is to maximize the fitness.

Fitness Convergence - A termination method that stops the evolution when the fitness is deemed as converged. Two filters of different lengths are used to smooth the best fitness across the generations. When the smoothed best fitness from the long filter is less than a user-specified percentage away from the smoothed best fitness from the short filter, the fitness is deemed as converged and the evolution terminates.

Population Convergence - A termination method that stops the evolution when the population is deemed as converged.
The population is deemed as converged when the average fitness across the current population is less than a user-specified percentage away from the best fitness of the current population.

Gene Convergence - A termination method that stops the evolution when a user-specified percentage of the genes that make up a chromosome are deemed as converged. A gene is deemed as converged when the average value of that gene across all of the chromosomes in the current population is less than a user-specified percentage away from the maximum gene value across the chromosomes.

HyperVolume Convergence - HyperVolume Convergence is a termination method for Multi-Objective Optimization in which the optimal solution set is measured by HyperVolume of the Pareto-Optimal Front. Evolution is stopped when the HyperVolume stabilizes.

Greedy Search & Back Elimination - Greedy Search is a termination method for AttributeSelection. When the Greedy Search algorithm is used, the evolution terminates immediately when adding a single input to the previous input set does not improve the fitness. When Back Elimination is used, the evolution terminates when removing a single input from the previous input collection leads to a worse fitness. A variation of the Greedy Search and Back Elimination is also implemented that keeps an elite pool with the size defined as EliteSize. When all solutions in the elite pool are checked using above criteria, the search is terminated.