That’s why in NLP we use term frequency over inverse document frequency. It gives you a measure of common uncommon things are.
Wonder how you’d implement that in a heat map. Just call each pixel a document and see where it takes you?
Wonder how you’d implement that in a heat map. Just call each pixel a document and see where it takes you?
A tf*idf matrix could be applied to the star-feature matrix too. Document = github repo. Term = name of user who starred it.
THUS, users who overstar are simply less important for computing similarities.
This would mitigate the phenomenon of massively popular github repos being clustered together because of folks who blithely star the most well known stuff.