3 ms·
I thought the article itself did a good job of explaining it. Backpropagation is just a special case of one way to do AD, apparently. AD prescribes a method of
by whatshisface 8y ago
I thought the article itself did a good job of explaining it. Backpropagation is just a special case of one way to do AD, apparently. AD prescribes a method of taking derivatives, and if you use it on a neural network, there you go, backpropagation.