Backpropagation
Linear Digressions
00:00
The Magic of Backpropagation
Backpropagation is sort of taking the errors and then propagating them backwards through your neural net. And you asymptotically, hopefully, if it's converging, it will approach the set of weights that are optimized over the whole neural net. These days, they're actually, they're neural nets are the state of the art for things like image recognition, also speech recognition, things like that.
Play episode from 08:57
Transcript
