Could you help me point out where the embedding vectors are updated when they are seemingly…

…: due to the fact that gradients bypass the embedding, we use a dictionary learning algorithm which uses an l2 error to move the embedding vectors eᵢ towards the encoder output:
Understanding Vector Quantized Variational Autoencoders (VQ-VAE)
554
11
Shashank Yadav
gordonlim
·Follow
Jun 7, 2021
--
Could you help me point out where the embedding vectors are updated when they are seemingly detached from the model?
--
--
Written by gordonlim2 Followers
·2 Following
No responses yet
Help
Status
About
Careers
Press
Blog
Privacy
Terms
Text to speech
Teams