Another optimization algorithm that leverages the first derivative or gradient of a function (Gradient Descent) is also widely used, and it determines high quality descent directions at a cost that is very often cheaper than the zero-order optimization algorithms.