Ilya Sutskever explains the origins of deep learning
The Robot Brains Podcast
00:00
Deep Learning by Hessian Ectimization
The first really piveral moment was when james martins has written a paper calledh deep learning by hessian ectimization. And it was the first time anyone has shown that you can train deep networks end to end from supervised data. The thing that' really clipped for me is, when i saw this, these results, wif e hessin free of optimise e realize, wat a second? So we can actually program those things. It means that if you have a neura letwork wich is pretty large, then there exist some barometers which achieve results on vision. Now, if only there was a data set which we could train from and tan
Transcript
Play full episode
