While learning Deep Learning using regression (Euclidean Loss), I stumbled on a question: Back-propagation minimizes the cost wrt the gradient descent and regression minimizes the cost of prediction, are they basically the same? Can anyone help me understand the difference?