Whereas @jameszou.bsky.social's approach to contrastive PCA has a parameter alpha (that can be difficult / impossible to choose), maximizing the RQ (we call this ρPCA) is straightforward, and produced much better results. We show this with many examples. 4/
