# 2020-04-06 | IA | Perceptron

###### tags: `swift` `ia` `gobelins`

## Import

import Foundation

## Class Perceptron

### Variables et constructeur

class Perceptron {

    var weights:[Double]
    var bias:Double = 0.01

    init(nbInput:Int) {
        weights = []
        for i in 0..<nbInput {
            weights.append(0.0) 
        }
    }
}

### Fonction Predict

    func predict(nbInput:[Double]) -> Double {
        var result: Double = 0.0
         for i in 0..<nbInput.count {
            result += nbInput[i] * self.weights[i] 
        }

        var sumOfProducts:Double = result + bias
        // print (sumOfProducts)
        return activate(sum:sumOfProducts)
    }

### Fonction Activate

func activate(sum:Double) -> Double {
        if sum <= 0 {
            return 0
        }
        else {
            return 1
        }
    }

### Fonction Train

    // dataSet = tableau de tuple
    // 1er => données
    // 2eme => result
        // lr = learningRate
    func train(dataSet[([Double],Double)],epoch:Int,lr:Double){
        for ep in 0..<epoch {
            for data in dataSet {
                let pred = self.predict(nbInput:data.0)
                let error = data.1 - pred

                var inputError = [Double]()
                for entry in data.0 {
                    inputError.append(entry * error)
                }
                
                // Mettre à jour les poids
                for i in 0..<inputError.count {
                    weights[i] += inputError[i] * lr
                }

                // Mettre à jour le biais
                bias += error * lr
            }
        }           
    }

## Call

let perceptron = Perceptron(nbInput:3)
print("Before training :")
print(perceptron.predict(nbInput:[1.0,0.3,4.0]))
print(perceptron.predict(nbInput:[0.0,0.5,1.0]))

// Création du dataset
let dataset = [([1.0,0.3,4.0],0.0),([0.0,0.5,1.0],0.0)]

// Training
perceptron.train(dataSet:dataset,epoch:6,lr:0.01)

// Test
print("After training :")
print(perceptron.predict(nbInput:[1.0,0.3,4.0]))
print(perceptron.predict(nbInput:[0.0,0.5,1.0]))

## Result

Before training :
1.0
1.0
After training :
0.0
0.0