Eric F. Lock
@elock.bsky.social
Biostatistician at University of Minnesota
For estimating low-rank signal in a matrix:
-Nuclear norm (NN) penalization over-shrinks when the signal-to-noise ratio is high 🙁,
-Hard thresholding (HT) the singular values over-fits when the signal-to-noise ratio is low 😲,
-But an empirical variational Bayes (EVB) approach is just right 😎.
-Nuclear norm (NN) penalization over-shrinks when the signal-to-noise ratio is high 🙁,
-Hard thresholding (HT) the singular values over-fits when the signal-to-noise ratio is low 😲,
-But an empirical variational Bayes (EVB) approach is just right 😎.
November 19, 2024 at 3:37 PM
For estimating low-rank signal in a matrix:
-Nuclear norm (NN) penalization over-shrinks when the signal-to-noise ratio is high 🙁,
-Hard thresholding (HT) the singular values over-fits when the signal-to-noise ratio is low 😲,
-But an empirical variational Bayes (EVB) approach is just right 😎.
-Nuclear norm (NN) penalization over-shrinks when the signal-to-noise ratio is high 🙁,
-Hard thresholding (HT) the singular values over-fits when the signal-to-noise ratio is low 😲,
-But an empirical variational Bayes (EVB) approach is just right 😎.