The Inside View cover image

5. Charlie Snell on DALL-E and CLIP

The Inside View

CHAPTER

Optimize the Image Given the Latent Vector

This is the evidence lower bound or usually pronounced as elbow. Basically this is the formula that's used to optimize like variational auto encoders and so these techniques are sort of instantiations, discrete versions of a Variational Auto Encoder VQVAE, DVAE. So what it does is it reparameterizes. The initial objective is to maximize the likelihood of the image given the latent vector. And so this sort of makes that tractable.

00:00
Transcript
Play full episode

Remember Everything You Learn from Podcasts

Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.
App store bannerPlay store banner