The paper derives a Riemannian update rule for the spectral factors of a positive-definite preconditioner, making arbitrary matrix roots fast and numerically stable for low-precision NN training.
Information geometry and its applications, volume 194
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
stat.ML 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
support 1representative citing papers
citing papers explorer
-
Spectral-factorized Positive-definite Curvature Learning for NN Training
The paper derives a Riemannian update rule for the spectral factors of a positive-definite preconditioner, making arbitrary matrix roots fast and numerically stable for low-precision NN training.