EngineParameters Class

EngineParameters

A class containing configuration information for the genetic engine. An object of this type is passed into the constructor of the engine.

Fields

Name	Data Type	Description	Typical Value	Default Value
PopulationSize	int	The number of candidate solutions in a generation	depends on the problem, maybe 200 - 1000	200
IsLowerFitnessBetter	bool	Indicates if a lower fitness score is desired	depends on the problem	true
SelectionStyle	enum	Tourney has the fastest performance but is more random (meaning, a lucky candidate with a bad fitness score can be easily selected), although that can be adjusted by increasing the TourneySize parameter. RouletteWheel is fitness-proportionate, but has slower performance. Ranked adjusts the fitness scores from their raw values into a ranking (to help smooth out situations where one candidate may be vastly better than others), and then uses Roulette Wheel to do the selection.	your choice	Tourney
TourneySize	int	When SelectionStyle is Tourney, specifies the number of candidates to randomly select, using the best of those	3 - 10, depends on population and how much you want to lean towards good candidates	4
CrossoverRate	double	A number from 0 to 1 that indicates how often crossover should happen. If crossover doesn't happen, the parent candidates are simply passed unaltered into the next generation	0.95, but it could be 1.0 if you're using elitism, which does roughly the same thing	0.95
MutationRate	double	Chance (from 0 to 1.0) that a candidate solution will be mutated, per generation	0 - 0.2	0.01
ElitismRate	double	A number from 0 to 1.0 that indicates what percentage (expressed as a decimal) of the population (sorted by fitness) should be passed unaltered into the next generation	0.05 - 0.20	0.10
RandomTreeMinDepth	int	When creating a random expression tree for generation 0 or mutation, the minimum tree depth desired	4 - 6	4
RandomTreeMaxDepth	int	When creating a random expression tree for generation 0 or mutation, the maximum tree depth desired	6 - 12	7
MinGenerations	int	The minimum number of generations to process, unless terminated by user	20 - 30	25
MaxGenerations	int	The maximum number of generations to process, unless terminated by user	50 - 100	75
StagnantGenerationLimit	int	How many generations to process without an increase in average fitness, before terminating	2 or 3 for quick tests, 10+ once you want the best solutions	12