Multi-class support ?

[gg4u]

Hi @marathomas !
I am looking at your repo, and a have a question:
does it support multi-class support ?

E.g. in the examples you look for clusters against 1 labels each: AGG, etc.

I would like to see if I can cluster against more : for each wav, there are a list of Labels [L1, L2, L3]

Other approachs I thought:

1. create a string:
   wav;"L1_L2_L3"

2. create a UMAP distinct for each class, and then pickup the coordinates for a point in each of the embedding, and average

But not sure if they still would make sense, mathematically, in reflecting the original structure.

Can you please share your opinion and point of view?
MSs student in data-science. Thanks!!