returning similar users after training (ranking an embedding)

apologies if there is a super simple answer to this that I have missed, but given a successful training of a recommender (e.g. with movielens), presumably it would not be too hard to return a list of similar users given an individual user id? I've being using tensorboard to get a visualization of the space of users. Here's an example from my own data set:

tensorboard itself allows one to specific a user id and get a ranked list of other users based on how close they are in the learned embedding space.

in a super simple slimmed down version with an embedding of only two dimensions we can see that each user id is represented as a 2-dimensional vector within the user_model:

[ 0.016, -0.027] ==> UNK
[ 0.454, -0.095] ==> 123456
[-0.44 ,  0.132] ==> 456788

so presumably there would be a relatively straightforward operation to rank all other users in order of their similarity to a given user, according to, say, euclidean distance in the n-dimensional space?

I'm sure I can do it element wise and then sort, but just wondering if there's an existing quick method as part of recommenders, rankings or embeddings?

Related?

• https://cloud.google.com/architecture/building-real-time-embeddings-similarity-matching-system
• https://stackoverflow.com/questions/43357732/how-to-calculate-the-cosine-similarity-between-two-tensors