As good as Ian Goodfellow's work has been, I think I have to disagree with him on what is part of the back propagation algorithm. Rumelhart et. al 1986 also des

As good as Ian Goodfellow's work has been, I think I have to disagree with him on what is part of the back propagation algorithm. Rumelhart et. al 1986 also describe it is a single algorithm from start to end. It is true you can use other things than stochastic gradient descent though, Mitchell also points this out by saying their algorithm was a backpropagation example using gradient descent.