Classic Perceptron implementation

main.cpp

#include <iostream>
#include <vector>
#include <cmath>

using namespace std;

// Step activation function
int activation(double x) {
    return 1.0 / (1.0 + std::exp(-x));
}

// Perceptron training
void trainPerceptron(vector<vector<double>> X, vector<int> y,
                    double &w1, double &w2, double &b,
                    double alpha, int epochs) {
    for (int epoch = 0; epoch < epochs; epoch++) {
        for (int i = 0; i < X.size(); i++) {
            double x1 = X[i][0], x2 = X[i][1];
            double z = x1 * w1 + x2 * w2 + b;
            int y_pred = activation(z);

            // Update weights and bias
            w1 += alpha * (y[i] - y_pred) * x1;
            w2 += alpha * (y[i] - y_pred) * x2;
            b += alpha * (y[i] - y_pred);
        }
    }
}

int main() {
    // AND training data
    vector<vector<double>> X = {{0,0}, {0,1}, {1,0}, {1,1}};
    vector<int> y = {0, 1, 1, 1};

    double w1 = 0.0, w2 = 0.0, b = 0.0; // Initialization
    double alpha = 0.1; // Learning rate
    int epochs = 10000; // Number of training epochs

    trainPerceptron(X, y, w1, w2, b, alpha, epochs);

    // Test the trained model
    cout << "0 OR 0 = " << activation(0*w1 + 0*w2 + b) << endl;
    cout << "0 OR 1 = " << activation(0*w1 + 1*w2 + b) << endl;
    cout << "1 OR 0 = " << activation(1*w1 + 0*w2 + b) << endl;
    cout << "1 OR 1 = " << activation(1*w1 + 1*w2 + b) << endl;

    return 0;
}