dyashkir / GPSwift

Simple Genetic programming in Swift (evolving programs to solve problems)

GPSwift

genetic programming in Swift

Basic Genetic Programming using Swift programming language, work in progress. Suggestions, ideas, PRs are welcome

##Install

Clone the repo ;)

Sample

Fit the X^2

var functionArray = [GPFunction]()

functionArray.append((+, "+"))
functionArray.append((-, "-"))
functionArray.append((*,"*"))

var leafs = [Leaf]()
leafs.append(Leaf())

struct SquareTrainer : GPTrainer{
    
    let train = [(1.0, 1.0), (2.0, 4.0), (3.0, 9.0), (10.0, 100.0)]
    
    func fitness(forProgram: ProgramTreeNode, eval: (ProgramTreeNode) -> Double, leafs: [Leaf]) -> Double {
        var score = 0.0
        train.forEach({ t in
            leafs[0].value = t.0
            let result = eval(forProgram)
            let error = (t.1-result)
            score += pow(error/t.1, 2.0)
        })
        return score
    }
}

let sq = SquareTrainer()

var run = GPRun(functions: functionArray, leafs: leafs, trainer: sq, initialDepth: 4, numberOfGenerations: 30)
run.start()

let maxPrg = run.currentGeneration?[0]

print("Best is:")

print(String(describing: maxPrg))

Handwriting number recognition

http://yann.lecun.com/exdb/mnist/ http://pjreddie.com/projects/mnist­in­csv/