Gradient Descent is used in machine learning to minimise a given cost function by iteratively adjusting the parameters of a model.