Latent Space: The AI Engineer Podcast cover image

ICLR 2024 — Best Papers & Talks (ImageGen, Vision, Transformers, State Space Models) ft. Durk Kingma, Christian Szegedy, Ilya Sutskever

Latent Space: The AI Engineer Podcast

NOTE

Optimizing VAEs through Amortization and Reparameterization

Using amortization, a framework of variational autoencoders introduced the idea of an inference model to approximate the true posterior, enabling fast inference over latent variables. By optimizing the evidence lower bound with respect to both the inference model and the true posterior using reparameterization, the algorithm achieves quick inference and training while optimizing a bound on log likelihood.

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