I'm tying to learn about recommendation systems recently. I have some deeplearning background so I focused more on machine learning based methods for recommendation systems. I see that a lot of paper directly train an embedding to represent the user (or part of the user). It is quite confusing to me since I believe that in real world, for big company like amazon or netflix. There will be new users every day. It's impossible to retrain the whole model to get embeddings for new users. So in real world, how do they deal with new users?
I have the same question for Matrix Factorization method. Is there any good source for answering those questions?