parallelismConfig

Creates ParallelismConfig with ParallelismConfigScope and apply it to the current CellularConfig.

Cellular Parallelism is based on the idea of dividing cell evolutionary strategies into independent processes that can be executed in parallel mode. This type of parallelism is more efficient than Panmictic Parallelism.

Base evolveCells stage function support Cellular Parallelism by default using process functions with ParallelismConfig.workersCount limit - it's safe to use!

NOTE results for CellularType.Asynchronous in parallel mode may be unpredictable (even a randomSeed set will not guarantee a repeatable result) since child chromosomes may replace parents at different rates depending on the operation of independent coroutines. However, this also introduces an element of randomness to evolution, which can have a positive effect on the entire process. CellularType.Synchronous is safe for parallelism because parallel evolutionary processes occur on an unchangeable parent population.

Example:

cGA {
    // configure specific params of CellularGA

    // configure Cellular parallelism
    parallelismConfig {
        workersCount = 5 // max count of parallel coroutines
        dispatcher = Dispatchers.Default
    }

    evolve {
        // evolving cells stage parallelism available
        // ParallelismConfig.workersCount limit by default
        evolveCells {
            selTournament(size = 3)
            cxOnePoint(chance = 0.8)
            mutFlipBit(chance = 0.1, flipBitChance = 0.01)
            evaluation()
        }

        stopBy(maxIteration = 50) { bestFitness == 100 }
    }
}

See also