Karl Rohe: New paper!
For the last century, we’ve misunderstood something fundamental about unsupervised learning.
This paper fixes it. 🧵
13 replies, 941 likes
Karl Rohe: Varimax with PCA has so many secrets
Muzhe Zeng (@muzheZ) and I found a big one in our recent paper
In this thread, I want to describe some more varimax secrets that we use to analyze data in my lab
3 replies, 145 likes
Peter Ralph: Use PCA? Check this out.
1 replies, 66 likes
Karl Rohe: We show how sparsity resolves the rotational invariance of factor analysis.
It gets better.
We show that PCA + Varimax estimates a huge class of "semi-parametric" models: SBMs, topic models, NMF, ICA, etc
Filling seminar slots? I’d love zooming to you
9 replies, 56 likes
alex hayes: very cool new work from my lab!
3 replies, 48 likes
alex peysakhovich 🤖: This is awesome. Varimax (and other "interpretability maximizing" rotations) are such an important but underrated topic in modern ML.
0 replies, 20 likes
Dixie Leonard Appreciation Society: This is a really cool paper!
1 replies, 5 likes
Dr. Max Halupka 🏳️🌈: Ftw
0 replies, 2 likes
Karl Rohe: One of the biggest secrets of PCA is that if you have "radial streaks" in your components, then you should try rotating with varimax.
here is a thread on a paper that describes this.
1 replies, 0 likes
Found on Aug 05 2020 at https://arxiv.org/pdf/2004.05387.pdf