Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Thanks for posting this.

I work in data science and I haven’t seen anybody claiming that PCA, an unsupervised technique, is a good classification (supervised) technique. I think this is the framing that the author is pursuing.

That said, some candidates might mention clustering because it’s easy to understand and you can apply an action (this group is high but also good customers so they get a special treatment).



I think PCA plus nearest neighbor was one of the earliest effective face recognition algorithms aka the eigenfaces method.


Finding rotations that maximize variance is not that far from euclidean based clustering.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: