19 Jul 2018

Sooooo clustering algorithms!

Silly old brain was blanking out and could only think of KMeans and DBSCAN, so did a little google.

I think sklearn is almost like a ‘cheatsheet’ (sort of) lol: most things you need, you got it there!

check this out:

For some next level stuff - if you have a gazillion (ok maybe like 80) features, you can use auto-encoders to reduce your features to fewer dimensions (say, 10?) and then use traditional clustering algorithms.

Auto-Encoders are like PCA on steroids - PCA is restricted to linear maps, while you can use non-linear functions for AEs.

For more deets, check this out: