ml-from-scratch / src /LinearRegression.java
Eric-Yyvuw's picture
Upload src/LinearRegression.java with huggingface_hub
989d20b verified
Raw History Blame Contribute Delete
3.59 kB
/**
* Linear Regression with Gradient Descent — CS401
* From scratch, no ML libraries.
*/
public class LinearRegression {
// Training data: Living area (m²) → Price (€k)
static double[] X = {50, 60, 70, 80, 90, 100, 110, 120, 130, 140};
static double[] Y = {150, 180, 210, 240, 270, 300, 330, 360, 390, 420};
static int n = X.length;
/** h(x) = theta0 + theta1 * x */
static double predict(double x, double theta0, double theta1) {
return theta0 + theta1 * x;
}
/** J(theta0, theta1) = (1/n) * sum((h(x_i) - y_i)^2) */
static double computeLoss(double theta0, double theta1) {
double total = 0;
for (int i = 0; i < n; i++) {
double yPred = predict(X[i], theta0, theta1);
total += Math.pow(yPred - Y[i], 2);
}
return total / n;
}
/** Gradient Descent */
static double[] gradientDescent(double lr, int iterations) {
double theta0 = 0;
double theta1 = 0;
for (int iter = 0; iter < iterations; iter++) {
double grad0 = 0;
double grad1 = 0;
for (int i = 0; i < n; i++) {
double yPred = predict(X[i], theta0, theta1);
double error = yPred - Y[i];
grad0 += error;
grad1 += error * X[i];
}
grad0 /= n;
grad1 /= n;
theta0 -= lr * grad0;
theta1 -= lr * grad1;
if (iter % 200 == 0) {
double loss = computeLoss(theta0, theta1);
System.out.printf("Iter %4d | Loss: %10.2f | theta0: %.4f | theta1: %.6f%n",
iter, loss, theta0, theta1);
}
}
return new double[]{theta0, theta1};
}
public static void main(String[] args) {
System.out.println("============================================================");
System.out.println("Linear Regression — Gradient Descent (Java)");
System.out.println("============================================================");
System.out.printf("Training data: %d examples%n", n);
System.out.print("X (area m²): ");
for (double x : X) System.out.print((int) x + " ");
System.out.println();
System.out.print("Y (price €k): ");
for (double y : Y) System.out.print((int) y + " ");
System.out.println();
System.out.println();
double[] theta = gradientDescent(0.00001, 2000);
double theta0 = theta[0];
double theta1 = theta[1];
System.out.println();
System.out.println("============================================================");
System.out.printf("Result: h(x) = %.2f + %.4f * x%n", theta0, theta1);
System.out.printf("Final loss: %.2f%n", computeLoss(theta0, theta1));
System.out.println("============================================================");
// Predictions
System.out.println();
System.out.printf("%10s | %12s | %14s%n", "Area (m²)", "Actual (€k)", "Predicted (€k)");
System.out.println("-".repeat(42));
for (int i = 0; i < n; i++) {
double yPred = predict(X[i], theta0, theta1);
System.out.printf("%10.0f | %12.0f | %14.2f%n", X[i], Y[i], yPred);
}
// Predict new value
System.out.println();
double newArea = 150;
double predictedPrice = predict(newArea, theta0, theta1);
System.out.printf("New prediction: %.0f m² → %.2f €k%n", newArea, predictedPrice);
}
}