AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Isn't a Pair of Images of Enything Just an Image?
What i believe in now is a sort of cross between gans and boltson machines. The idea of a mark off chain to generate the negative data, i think it's just too cumbersome. I think we need otherways of generatin negative data. And by finding structure that should only be there if it's real data, that's the sort of main goal. A paper with the wake sleep algaism that i published in two thousand and six with simonsandaro and u i t in your computation. We have something, but that doesn't use back prop it manages to learn well without back prop it uses contrasted wake sleep.