Ali Ghodsi, Deep Learning, Variational Autoencoder, VAE, Performer, Fall 2023, Lecture 15

preview_player
Показать описание
This lecture primarily focuses on Variational Autoencoders (VAE), beginning with an overview of variational inference and the computation of the Evidence Lower Bound (ELBO) for VAE, which starts at minute 24. Prior to this, the video discusses Performers, a variation of Transformers, based on the paper "Rethinking Attention with Performers."

Keywords: Variational Autoencoders, VAE, variational inference, ELBO, Performers, Transformers.
Рекомендации по теме
Комментарии
Автор

thank you, professor! The explanation is very intuitive.

longh
Автор

رائع جدا ..الله يعطيه العافيه ويسعده ويوفقه ويوفق عياله ان كان عنده عيال

Rasha-tcbl
Автор

Thank you professor. Very precise presentation!

mdmamunurrashid
Автор

at 28:00, shouldn't the reconstruction loss be " || x - x_ || = || x - UU^Tx || " to be minimised as opposed what's given in the video i.e || x_ - UU^Tx || ?

prateekpani