Episode 13: Jonathan Frankle, MIT, on the lottery ticket hypothesis and the science of deep learning
Generally Intelligent
00:00
Prue Networks of Initialization - Is This a Good Technique?
Ono: A lot of methods that have been proposed in literature over the past few years claim they can prue networks of initialization and do really well. He says he started trying to implement these techniques as base lines for his research, and found a lot of weird things. Ono: We need a little more rigor. And we should be open to the fact that people can say what's wrong with their ideas.
Transcript
Play full episode
