Study Deep Learning from scratch in Java.

※Caution This article is low level and has a lot of self-satisfaction that I am writing for myself, just starting to read this book The content may be incorrect or the code may be strange. Also, I don't explain terms so much, so I think it's meaningless to people who don't know machine learning. I will update this article gradually.

(Single layer) Perceptron implementation

Perceptron: A type of algorithm. The algorithm that was the basis of the neural network. A step function is used as the activation function. With a single layer, only a linear region can be expressed.

** Express AND, OR, NAND by passing the bias and the weight of each input signal to the constructor. ** **

public class PerceptronSample {

    private double b, w1, w2;

    PerceptronSample(double b, double w1, double w2){
        this.b = b;
        this.w1 = w1;
        this.w2 = w2;
    }

    int perceptron(int x1, int x2){
       double tmp = (x1*w1 + x2*w2) + b; //Sum
        return tmp > 0 ? 1: 0;
    }




    public static void main(String[] args) {

        PerceptronSample perceptronSample = new PerceptronSample(-0.5, 0.3, 0.3); 
        //bias,Input 1,Input 2
        //This time AND circuit


        System.out.println("" +
                perceptronSample.perceptron(0, 0) +
                perceptronSample.perceptron(1, 0) +
                perceptronSample.perceptron(0, 1) +
                perceptronSample.perceptron(1, 1)
        );

    }

}

The result is 0001.

Express XOR with multi-layer perceptron.

A deeper layer of perceptron. It can solve problems that cannot be solved by a single-layer perceptron.

XOR is expressed by NAND, OR and AND.

public class MultilayerPerceptronSample {

    private double b, w1, w2;

    MultilayerPerceptronSample(double b, double w1, double w2){
        this.b = b;
        this.w1 = w1;
        this.w2 = w2;
    }

    int perceptron(int x1, int x2, MultilayerPerceptronSample m1, MultilayerPerceptronSample m2){

        if (m1 != null && m2 != null){
            int x[] = {x1, x2};
            x1 = m1.perceptron(x[0], x[1], null, null);
            x2 = m2.perceptron(x[0], x[1], null, null);
            double tmp = (x1*w1 + x2*w2) + b;
            return tmp > 0.0 ? 1 : 0;
        }else {
        double tmp = (x1*w1 + x2*w2) + b; //Sum

        return tmp > 0.0 ? 1 : 0;
        }
    }






    public static void main(String[] args) {

        MultilayerPerceptronSample AND = new MultilayerPerceptronSample(-0.5, 0.3, 0.3);
        MultilayerPerceptronSample NAND = new MultilayerPerceptronSample(0.5, -0.3, -0.3);
        MultilayerPerceptronSample OR = new MultilayerPerceptronSample(-0.3, 0.5, 0.5);




        System.out.println(" " +
                AND.perceptron(0, 0, NAND, OR)+
                AND.perceptron(1, 0, NAND, OR)+
                AND.perceptron(0, 1, NAND, OR)+
                AND.perceptron(1, 1, NAND, OR)


        );


    }

}