Gradient Descent: How Models Actually Learn

Behind every trained model is gradient descent — an algorithm that nudges model weights in the direction that reduces error. Small steps, millions of times.