The mathematical derivative is fundamental for optimizing machine learning algorithms. In deep learning, backpropagation gradient descent is widely used for training feedforward neural networks. Explain how derivative is applied in backpropagation gradient descent to optimize neural networks.