3 ms·
From my understanding the vectors don't really matter so much. We care more about the subspace they span (because we want to project our data into it). So any r
by gubbrora 8y ago
From my understanding the vectors don't really matter so much. We care more about the subspace they span (because we want to project our data into it). So any rotation is as fine as any other.
- gubbrora 8y agoOn further thought I think this is wrong. Large rotations can cause coefficients to become negative, so there are constraints.