I'm trying to develop a similarity matrix that downweights any item that is similar to many things. Does anyone have experience with this? from r/datascience