Backpropogation

Backpropogation is the way that deep learning neural networks learn. The whole idea behind it is to try to figure out how to nudge each layer so that the cost function is minimized, starting from the last layer and going back.

See Also