Just for fun, this project allows you to use some very simple genetic algorithm code to search for solutions. It is very basic, and just supports Int/Double/Boolean genes at the moment, and just uses simple gaussian mutation for numerics and random flips for booleans.
To use it, simply define a case class that represents your encoded solution - the attributes need to be of the type Gene[A]
For example:
case class ThreeNumbers(x: IntGene, y: IntGene, z: IntGene)
You can then provide an instance of that solution (the instance allows additional configuration to be set, such as max/min for numerics) along with an errorRate function (function that given a solution as input returns a score that represents the rate of error for how it performs)
For example, the below configuration uses the above defined ThreeNumbers case class and attempts to search for a solution that such that the product of the three numbers is as close as possible to 10,000
EvolutionFactory(
candidate = ThreeNumbers(IntGene(min = 0, max = 100), IntGene(min = 0, max = 100), IntGene(min = 0, max = 100)),
populationConfig = PopulationConfig(50),
evolutionConfig = EvolutionConfig(50),
errorRate = (c: ThreeNumbers) => Math.abs(10000.0 - (c.x.value.get * c.y.value.get * c.z.value.get))
).run