The Importance of Scaling an Auto Encoder

It's more efficient to sample from the trained auto encoder's reconstruction of like data from the original training set than it is to Sample from the training set. The reconstructions by an auto encoder can be more informative about the structure of the world than the world itself and this only works when we're talking about a relatively small amount of sampling. A good work of art is something thatlike is more informative about the world than theworld is informative about the world. We don't have infinite sampling power like anymore because of central limit theorem, he says.

Transcript

Play full episode

The AI-powered Podcast Player

Save insights by tapping your headphones, chat with episodes, discover the best highlights - and more!

Get the app