-
The Spectrum Was Known to Be Biased. Why Did PCR Keep Inverting It?
Covariance estimation learned to correct the distorted high-dimensional sample spectrum; principal component regression kept inverting it. De-floored PCR removes the endogenous spectral floor before inversion — in the regime where it is large enough to matter yet cheap to remove.
-
Shrinkage Is Not a Universal Law
Why anisotropic prediction geometry can make expansion, rather than shrinkage, the right correction in overparameterized regression.